Government control and the value of cash
Table 3 reports the estimation results in Panel A and the marginal value of cash calculated for an average firm in Panel B.Footnote 5 The baseline regression, similar to Faulkender and Wang (2006), without government control included, is reported in column 1 of Panel A. The coefficient estimate shows that an extra unit (RMB) of cash increases shareholder wealth by RMB 1.62 if a firm has no cash holdings or debts. As reported in column 1 of Panel B, on average, the value of an additional RMB of cash holdings to shareholders in the mean firm is RMB 1.00. Our result is very similar to what is found by Faulkender and Wang (2006), who report a slightly lower marginal value of cash to shareholders of US $0.94. One potential reason for the difference is that firms in China, as in many emerging countries, regard financial constraints as one of their primary obstacles to funding profitable investments, given the country’s high speed of development (Cull et al. 2015). Therefore, cash is more valuable to shareholders in China than in the US due to operational considerations and precautionary motives (Faulkender and Wang 2006; Denis and Sibilkov 2009).
Table 3 Government control and the value of cash: Main results All control variables are significant with expected signs, consistent with the literature (Faulkender and Wang 2006; Dittmar and Mahrt-Smith 2007). Specifically, we find that the excess stock return is positively correlated with the change in profitability, the change in net assets, the change in interest expenses, the change in dividends and lagged cash holdings; and it is negatively associated with market leverage. The coefficients on the change in cash interacted with cash holdings and with market leverage are both significantly negative, supporting the view that the value of cash decreases as a firm holds more cash and debts. The results are plausible, since firms with little or no cash reserves tend to have costly access to external finance, thereby benefiting the most from additional liquid assets. Similarly, since the likelihood of default increases with the level of debts, the value of cash decreases with debts.
In column 2, government control, measured by government dummy, and its interaction with the change in cash are included. The coefficient on the interaction term ∆CASH × GOV is negative and statistically significant. Economically, an extra unit (RMB) of cash held in an average firm controlled by the government is valued RMB 0.25 less than that is held in a privately controlled firm. To address the concern of time-specific and time-invariant firm- or industry- level factors, we run the fixed effects regression controlling for year and industry or firm heterogeneity, as shown in columns 3 and 4, respectively. The results indicate clearly that our main finding continues to hold.
We then partition our sample into government-controlled and privately controlled firms, and conduct a subsample analysis as a comparison. The regression results and the difference of government- and privately controlled firms are shown in columns 5–7. The coefficient on ∆CASH is 1.48 and 1.91 in firms controlled by the government and by private owners, respectively. The difference (− 0.43) is statistically significant at the 5% level. Moreover, the significance and magnitude of the control variables vary across the two subsamples. Specifically, firm value, measured as the excess stock return, is less positively affected by a change in profitability or change in net assets and is more positively affected by a change in dividends and cash holdings in state-controlled firms compared to privately controlled ones. The negative effect of market leverage on the excess stock return is weaker in state-controlled firms than in privately controlled firms. Interestingly, the value of cash decreases less as market leverage increases in state-controlled firms than in privately controlled firms. One possible explanation is that the likelihood of default is low in firms controlled by the state due to the implicit bailout guarantee during distress.
Overall, the evidence presented in this section is in line with our first hypothesis, suggesting that the value of cash is discounted by investors for firms under the government control.
Robustness tests
Endogeneity concern
In the earlier analysis, we have attempted to mitigate the endogeneity concern using fixed effect regressions by controlling for year, industry, or firm heterogeneity, which may alleviate the omitted-variable bias. Our main result is interpreted based on the coefficients on the interaction between government control and the change in cash. Cash holdings may change substantially over time, but government control may stay relatively stable. Therefore, if endogeneity exists, it tends to affect the coefficient of government control rather than its interaction with the change in cash (Dittmar and Mahrt-Smith 2007). Nevertheless, we now conduct several additional analyses to further address the possible endogeneity concern, as shown below.
First, we use the lagged variable on government control. It is not likely that the lagged government control is endogenously determined with the current excess stock return. As shown in column 1 of Table 4, the main result remains qualitatively unchanged. Our result is also robust using the 2- and 3-year lagged variables on government control.
Table 4 Addressing endogenous concern Second, to control for observable differences in firm and industry attributes, we next perform the analysis based on a propensity-score matched sample. We first run a logit model regressing the likelihood of a given firm to be controlled by the government on firm size, firm age, book-to-market ratio, leverage, return on assets, and year and industry effects. Then, the likelihood (i.e. the propensity score) that a firm is government-controlled is estimated. Each observation in the government-controlled group is matched to an observation in the privately controlled group based on the nearest neighbor technique. We allow for replacement and require the difference in the propensity scores for each pair to be within 0.1% in absolute value. Our final sample comprises 19,594 firm-year observations, with 9797 firm-years being controlled by the government and the remaining 9797 firm-years by private owners.Footnote 6 The result of using the matched sample is shown in column 2 of Table 4. As evident, the negative impact of government control on the value of cash still holds.
Finally, we perform an analysis focusing on firms that have undergone through changes in their type of ultimate controlling from the state to private owners during the sample period, as such a shift may lead to a decrease in agency costs (or an increase in financial constraints) that may affect the value of cash. To validate our main finding, reported earlier, we expect to observe a positive impact of such a move on the value of cash. During our sample period, a total of 187 firms have experienced such transformation, and an indicator, Transfer, is constructed such that it is equal to one in the year when the firm is transferred to private ownership and afterwards, and zero in years preceding the transformation. We re-estimate Eq. (1) by replacing government control (GOV) with the indicator variable, Transfer, and report the result in column 3, which shows a significant increase in the value of cash following such a transformation from government control into private hands, verifying our main finding—that government control does reduce the value of cash.
Alternative measures of government control and the expected change in cash
To evaluate the sensitivity of our main finding, we re-estimate Eq. (1) using three alternative measures of government control. The first two, GOV1 and GOV2, are dummies with 10% and 20% cut-offs of ultimate control right, respectively, and the third, GOV_cont, is a continuous measure of government control using ultimate control right. The results are reported in columns 1–3 in Table 5.Footnote 7 We find that our main prediction related to government control remains unaffected: The coefficient on the interaction term ∆CASH × GOV is negative in all three columns, and is statistically significant at the 1% level for both GOV1 and GOV2, and at the 10% level for GOV_cont.
Table 5 Robustness checks: alternative measurement of key variables and additional sets of controls Moving to our key independent variable, the change in cash is defined as the unexpected change in cash holdings. The results reported thus far assume that the expected change in cash is equal to zero, and therefore, the change in cash is, in fact, the realized change in cash. Consistent with Faulkender and Wang (2006), we now conduct robustness checks using three alternative definitions to measure the expected change in cash, as shown below.
The first measure is the average change in cash of the corresponding benchmark portfolio over the year (Portf. Ave). It follows that if most firms in the same size and book-to-market portfolio change cash reserves over the year, such change should be reflected in the benchmark return already, and the excess return is one that is not revealed in the benchmark return. With respect to two other measures, we adopt two models from Almeida et al. (2004) to obtain the expected change in cash. In both cases, changes in cash are regressed on factors that represent sources and uses of cash. The unexpected change in cash is measured as the difference between the actual change in cash and the predicted change in cash from the models (i.e. residuals). The first model, ACW (1), is as follows:
$$\Delta CASH_{i,t} = \alpha_{0} + \alpha_{1} CF_{i,t - 1} + \alpha_{2} MB_{i,t - 1} + \alpha_{3} SIZE_{i,t - 1} + \varepsilon_{i,t}$$
(2)
where ΔCASH is the change in cash and cash equivalents, CF is cash flow, MB is market-to-book ratio and SIZE is firm size. All variables are deflated by the lagged market value of equity, except for the market-to-book ratio (MB) and firm size (SIZE).
The second model, ACW(2), adds capital expenditures (CAPEX), change in net working capital (ΔNWC) and change in short-term debt (ΔSTD), all lagged deflated by the lagged market value of equity, as additional explanatory variables. The equation is shown below:
$$\begin{aligned} \Delta CASH_{i,t} & = \alpha_{0} + \alpha_{1} CF_{i,t - 1} + \alpha_{2} MB_{i,t - 1} + \alpha_{3} SIZE_{i,t - 1} + \alpha_{4} CAPEX_{i,t - 1} + \alpha_{5} \Delta NWC_{i,t - 1} \\ & \quad + \alpha_{6} \Delta STD_{i,t - 1} + \varepsilon_{i,t} \\ \end{aligned}$$
(3)
The results are shown in columns 4–6 of Table 5. The expected change in cash is measured as the portfolio average in column 4, ACW (1) in column 5 and ACW (2) in column 6. Overall, using different measurements for the expected change in cash generates nearly identical results to those reported in Table 3, consistent with Hypothesis 1, that government control reduces investors’ valuation of cash held by firms.
Controlling for corporate governance
To reduce the concern about potential correlated omitted variables, we further examine whether our main finding is sensitive to the inclusion of corporate governance variables. Three proxies are considered from the aspects of excess control rights, institutional ownership and analyst coverage. First, the ratio of control rights to cash-flow rights by the ultimate controllers is used to measure excess control rights (Lemmon and Lins 2003; Xu et al. 2016). We use 10% as the cut-off point to determine effective control at the ultimate level (Claessens et al. 2002). Large excess control rights entrench the controlling shareholders and give them the ability to tunnel, while small cash-flow rights limit controlling shareholders’ wealth losses from the tunneling. Thus, firms with more excess control rights are worse governed.
Second, we follow Firth et al. (2016) and define institutional ownership (INST) as the percentage of shares owned by domestic mutual funds and qualified foreign institutional investors (QFII). Compared with other institutional investors, such as banks, insurance and securities companies, domestic mutual funds and QFII have a higher exit threat and thus are more active and effective in monitoring. Finally, analyst coverage (ANACOV) is measured as the natural logarithm of one plus the number of analysts following the firm (Feng et al. 2016). Analyst coverage plays an important monitoring role in reducing earnings management and managerial expropriation (Yu 2008; Chen et al. 2015). Each of the three governance variables and their interactions with the change in cash is added into Eq. (1) one by one. The results are reported in columns 7–9 of Table 5. Consistent with prior literature, firms with low excess control rights, high institutional ownership and high analyst coverage have a higher value of cash. Significantly, the negative impact of government control on the value of cash still exists, which is above and beyond the impact of corporate governance.Footnote 8
Why does government control lead to a lower value of cash?
As discussed earlier, the negative impact of government control on the value of cash may arise from two possible channels: One is from the high agency costs associated with political objectives and managerial interests due to poor governance mechanisms, which we refer to as agency costs of political expropriation channel, and another is through low financial constraints inherited in firms under government control due to preferential access to credit provided by the government, which we refer to as financial constraints of the soft-budget effect channel. In this section, we investigate mechanisms through which the government control reduces the value of cash. Specifically, we examine how government affects a firm’s dissipation of cash holdings, financial constraints and external financing channels.
Government control and dissipation of cash holdings
According to the agency costs channel, cash holdings can be easily diverted by corporate insiders to be used in projects in accordance with political considerations, such as increasing employment rate and maintaining social stability and gaining national pride, rather than project merits. With poor monitoring systems, managers in government-controlled firms are entrenched and may use free cash flows to overinvest in empire-building (Jensen and Meckling 1976). In line with the free cash flow hypothesis, prior studies find that excess cash and week governance schemes lead to increases in capital expenditure (Iskandar-Datta and Jia 2014). The agency costs of political expropriation channel predict that firms under government control may increase the use of cash in subsequent investments, especially those with low investment efficiency (Jaslowitzer et al. 2016; Chen et al. 2017). In contrast, if the financial constraints channel holds, firms under government control may face a lower degree of financial constraints and have better access to external financing, such as bank loans, and they are less likely to use internal finance, such as cash, for investments (Denis and Sibilkov 2009). Therefore, the financial constraints channel suggests that government control is less associated with the use of cash holdings in subsequent investment.
We test this premise by following Gao et al. (2013) to estimate a logit regression with the dependent variable being the increase in investment (INCR_INV), a dummy variable that takes value of one if a firm increases investment in the next year, and zero otherwise. The impact of government control on the use of cash in subsequent investment is indicated by the coefficient on the interaction between cash and government control (CASH × GOV). The agency costs (or financial constraints) channel implies that firms under government control are more (less) likely to increase the use of cash in subsequent investment, that is, a positive (negative) coefficient on the interaction. Controlling variables include Tobin’s Q (Q), firm size (SIZE), cash flow (CF), book leverage (BLEV), dividend dummy (DIV), sales growth (SG) and return on asset (ROA). The results are shown in Table 6, where investment is measured as capital expenditure in column 1, and the sum of capital expenditure and acquisitions in column 2. As evident, the coefficient on the interaction term between cash and government control is positive and statistically significant in both columns, supporting the agency costs channel, that firms with government control disgorge more cash via investments compared with privately controlled firms.
Table 6 Disgorging cash via investment and investment efficiency Next, we investigate whether firms with government control use cash to invest in projects associated with corporate innovation. The rationale of conducting this test is that if firms with government control invest in projects for political objectives, cash is less likely to be employed in innovation activities, such as R&D. The agency costs of political expropriation predict a negative relationship between cash and R&D in firms under government control, while the financial constraints of the soft-budget effect channel predicts no such relationship. To test for this, we define the increase in R&D (INCR_RD) as a dummy variable that is equal to one if a firm increases R&D in the next year, and zero otherwise.Footnote 9 The result is shown in column 3, where the coefficient on the interaction term CASH × GOV is negative and statistically significant, confirming our conjecture that firms under government control are less likely to disgorge cash via innovation activities.
Finally, we test for the impact of government control on investment efficiency. The extant literature suggests that firms with government control may be obligated to invest in politically favored projects that are unprofitable, leading to investment inefficiency (Chen et al., 2011b). The agency costs channel suggests a negative impact of government control on investment efficiency, while the financial constraints channel does not. We follow Chen et al (2011b) and use the sensitivity of investment expenditure to investment opportunities (Tobin’s Q) as a measure of investment efficiency. Results are reported in columns 4 and 5 of Table 6. The results indicate that the coefficient on the interaction term, Q × GOV, is negative and highly significant when investment is measured as capital expenditure (column 4), the sum of capital expenditure and acquisitions (column 5). The negative impact of government control on the efficiency of investment further supports the agency costs of political expropriation associated with government control.
Taken together, the results reported in Table 6 support the agency costs of political expropriation channel in that the state uses corporate cash under its control to invest in projects that mainly achieve political objectives, with less concerns on innovation and investment efficiency, leading to a lower value of cash for firms controlled by the government than by private owners. Our finding is consistent with Chen et al. (2011b) and Chen et al. (2017), who demonstrate that the government control distorts firm investment behavior and harms investment efficiency. Similar, Jaslowitzer et al. (2016) find that state ownership restrains firms’ responsiveness to investment opportunities, and Boubakri et al. (2013a) contend that state ownership is negatively associated with corporate risk-taking.
Financial constraints and external financing channels
The evidence provided earlier suggests that the negative impact of government control on the value of cash is through agency costs of political expropriation. In this section, we conduct several direct tests showing that firms under government control may not be necessarily subject to the financial constraints of the soft-budget effect. Specifically, if the financial constraints channel is valid, government control should be expected to enhance corporate external financing ability and alleviate financial constraints. We measure the degree of financial constraints using cash flow sensitivity of cash, based on two models developed by Almeida et al. (2004),Footnote 10 and external financing using gross equity issuance and net debt issuance, based on the model adopted by Firth et al. (2012). The financial constraints channel suggests a negative impact of government control on cash flow sensitivity of cash and a positive impact of government control on equity and debt issuances. The results for the cash flow sensitivity of cash are presented in columns 1 and 2 of Table 7. It shows clearly that the coefficient on the interaction term between the government control and cash flow (GOV × CF) is not statistically significant in both cases, confirming our premise that firms under government control are not particularly subject to soft-budget on financial constraints compared to firms under private control. In terms of external financing for gross equity issuance and net debt issuance, shown in columns 3 and 4, respectively, the coefficient on government control is negative in both cases; statistically significant for gross equity issuance; and marginally significant for net debt issuance, implying that government control has, in fact, a negative impact on external financing. Overall, our results are consistent with recent literature, suggesting that government control does not necessarily enhance financing ability or mitigate financial constraints (Ben-Nasr et al. 2012; Firth et al. 2012; Borisova et al. 2015; Jaslowitzer et al. 2016).
Table 7 Government control, financial constraints and external financing channels Based on the evidence provided in this section, we conclude that the negative impact of government control on the value of cash is solely due to the severe agency costs related to the political expropriation, rather than financial constraints of the soft-budget effect.