The Constraint Effect on the Value of Cash Holdings of
US Firms in times of Crisis and Post-crisis
Abstract
This paper examines the marginal value of cash holdings before, during and after the crisis, conditional on financial constraints. This paper shows that the US firms generally experienced a decrease in cash holdings during the crisis and raised their cash positions afterwards as a response to the crisis. Furthermore, this paper confirms the cash effect on enhancing firm value. Contrary to the previous literature, this paper finds that the crisis effect and the post-crisis effect have little impact on the value of cash holding for both constrained and unconstrained firms. These results are consistent when the reporting lag is taken into account.
Key words: Cash holdings, Marginal value, Financial crisis, Financial constraints JEL classification: G01, G31, G32
Student name: Yun Si Student number: s2008769 Study program: MSc Finance Supervisor: Prof. dr. Bert Kramer
June 8, 2017
1. Introduction
How do shareholders evaluate the cash held by firms? Modigliani and Miller (1958) state that in a perfect capital market, any value-enhancing investments would be financed. In addition to that, firm value would increase exactly as much as the value of those investments regardless how the investments are funded. Hence, the level of cash holding is irrelevant. However, given that the capital markets are imperfect, at least transaction costs and taxes make firms are no longer indifferent between financing the investments with internal liquidity such as cash and with external funds. On one hand, holding cash as internal liquidity allows firms to avoid transaction costs, underinvestment and to reduce the likelihood of involving in financial distress costs (Opler, Pinkowitz, Stulz and Williamson, 1999 and Faulkender and Wang, 2006). On the other hand, holding cash comes at cost. The agency theory indicates that holding excess cash allows managers to achieve personal interests at the expense of shareholders. Excess cash holdings increases the probability that the managers would invest in potentially value-decreasing projects which are harmful to the firm value (Jensen, 1986 and Pinkowitz and Williamson, 2004).
Previous studies made great contributions to investigate what determines how much cash firms should hold and what value the market places on firms’ cash holdings. Opler et al. (1999) examine the determinants of cash holdings, and find that firms with strong growth opportunities and riskier cash flows are inclined to hold more cash. Faulkender and Wang (2006) empirically find that the value of one extra dollar cash held by firms should vary depending on how the extra cash is going to be used. The one extra dollar cash that is distributed to either equity holders or debt holders would have a value lower than $1. But for those firms with insufficient cash reverses, the value of one extra dollar cash would be higher than $1 when they need to raise capital from outside markets to invest in value-enhancing projects.
held by firms is more valuable during the crisis. That is, the crisis effect should result in a higher value of cash holdings. However, the post-crisis effect on the value of cash is seldom investigated in the previous studies. Firms, after experiencing such a shock, would increase their cash holdings, but the increase in cash holdings would conversely decrease the value of one additional dollar cash raised by firms. In other words, the post-crisis effect should give rise to a lower value of cash holdings.
Another strand of literature emphasizes that firms’ accessibility to external markets also has an impact on the value of cash holdings (Denis and Sibilkov, 2007, Duchin, Ozbas, and Sensoy, 2010 and Campello, Graham, and Harvey, 2010). Faulkender and Wang (2006) present that cash is more valuable for constrained firms since the higher cash holding increases the probability that the value-enhancing investment would not be bypassed. As a result, the greater constraint faced by the firm, the higher the value of cash holdings of the firm. In addition, the financially constrained firms should be affected more by the crisis than others. Campello et al. (2010) find that the constrained firms experienced a more severe crisis by reducing more employment and cutting more capital investment and market expenditures. Chang, Benson and Faff (2016) and Kahle and Stulz (2013) provide the counter-argument that the unconstrained firms should be affected more. This is because the unconstrained firms rely more heavily on credit, so that the reducing supply of credit during the crisis would lead to a more painful experience by the unconstrained firms.
As discussed above, there are incomplete conclusions on the crisis effect and the constraint effect on firms’ cash holdings. Additionally, little attention has been paid to the post-crisis effect on firms’ cash holdings. In this study, we therefore want to investigate the following research questions:
How did US firms change their cash position in response to the recent global financial crisis, and did firms increase their cash holdings for the post-crisis time period?
What is the crisis effect on the marginal value of cash holdings of US firms, and does a post-crisis effect on cash holdings exist?
cash holdings?
According to above research questions, this paper explores the crisis effect, the post-crisis effect and the constraint effect on the marginal value of cash holdings. The OLS regression models are applied on the basis of two panel datasets. One is the balanced sub-sample dataset and the other one is the unbalanced whole sample dataset. Specifically, I use data covering the period 2002 to 2015, with 2008-2009 being defined as crisis period and 2010-2015 being defined as post-crisis period. An alternative definition of crisis (post-crisis) period is applied according to the evolution of US firms’ cash positions. Three different classification criteria are used to construct the constraints dummy. The triple-interactions of cash with the crisis effect and the constraint effect, and the post-crisis effect and the constraint effect are created to capture how the value of cash would be assessed during the crisis and after the crisis for both the constrained firms and the unconstrained firms.
This study finds out that the changes in cash holdings positively affect the firm’s equity market performance, and it provides the evidence that US firms experienced a cash shortage at the start of the crisis, but they rapidly increased their cash positions as a response to the crisis. The cash-to-asset ratios of US firms, on average, sharply decreased in 2008. Due to the precautionary purpose and the bad economic conditions, US firms increased their cash holdings. The cash-to-asset ratios climbed up, peaked at 2010 and maintained a stable level in following years. In addition, because of the inconsistent results, I find that there is little evidence that the value of the extra cash holdings is different in times of crisis and post-crisis compared to the pre-crisis period. Simultaneously, Contrary to the previous studies, I find that the marginal value of cash holdings are likely to be same for both constrained and unconstrained firms. This indiscrimination holds for both pre-crisis, during-crisis and post-crisis periods. Overall, it indicates that the marginal value of cash holdings is less impacted by the joint effect of the crisis (post-crisis) effect and the constraint effect.
post-crisis effect on the marginal value of cash holdings vary depending upon the definition of the crisis period, and for financially constrained firms, the crisis effect and the post-crisis effect are unlikely to have an impact on the value of cash holdings.
The remainder of this paper is organized as follows. Section 2 reviews the previous studies related to this study. The methodology is introduced in section 3 along with the definitions of the crisis (post-crisis) period and the alternative approaches to classify the constrained and the unconstrained firms. Sample selection and data statistics are presented in section 4. Section 5 reports the main results of the regression models. Section 6 concludes the findings, points out the limitations of this study and suggests directions for future research.
2. Related Literature
2.1 The Motivations of Cash Holding
In a perfect capital market, Modigliani and Miller (1958) indicate that the firm can always costlessly raise capital for any value-enhancing investments. The frictionless capital market assumes that firm value is not influenced by the firms’ payout policy. Therefore, the firm’s choice between keeping retained earnings as cash holdings and paying out as dividend to their equity holders is irrelevant since they can certainly be financed when they need capital. In the case of perfect capital markets, firms seem to have no motivation to hold liquid assets such as cash on hand because the extra cash will not improve the firm value, and firms will simply prefer to distribute their earnings to shareholders to delight them.
Mauer, Sherman, 1998 and Miller, Orr, 1966). Firms, with frequent business with other parties, or with frequent demand to outside capital markets are induced to hold cash for the purpose of costs reduction. Opler et al. (1999) also explain that firms have a strong intention to hold cash because of precautionary motives, which suggests that the available cash held by firms can effectively help firms against future uncertainties. Firms with sufficient cash on hand are more confident to meet unexpected contingencies, to reduce financial distress costs, and are less likely to miss any value-enhancing investments. This motivation particularly holds for firms with volatile cash flows or limited access to external capital markets.
Another motivation that has been discussed by previous literature is the tax-based motivation. Pinkowitz, Stulz and Williamson (2013) and Foley, Hartzell, Titman and Twite (2007) explicitly argue that U.S. multinational cash holdings are partly a result of the tax cost associated with repatriating foreign income. The adverse tax consequence of repatriating the cash back to the domestic country provides the incentive for multinationals to hold their cash abroad. This argument is supported by Grubert (1998) and Grubert and Mutti (2001) who find that repatriations are sensitive to tax considerations, and that the multinationals are inclined to retain their earnings in low tax areas. Hines and Hubbard (1990) further indicate that this relationship not only exists between repatriation tax and firms’ cash holdings, but they also show evidence that the dividend payments by foreign subsidiaries are significantly influenced by repatriation tax. Nevertheless, this motivation on the basis of tax considerations may be only valid for relatively large firms who run their business in different countries, but not for small local firms, as the repatriation tax seems to be irrelevant with their local business.
with the interests of shareholders. Pinkowitz and Williamson (2004), Jensen (1986) and Harford (1999) show that the cash-rich firms are more likely to engage in value-damaging activities, simply as a result of the fact that the available excess cash allows managers to take additional risks or to invest in unnecessary projects. Even though the motivations of holding cash by firms is not the main question to be explored in this study, it still provides strong incentives for me to understand why firms want to hold cash and how precisely the firms’ holding behaviors can be explained by these motivations.
2.2 The Trends of Cash Holding
Bates et.al (2009) empirically find out that the average cash-to-assets ratios for United State firms experienced a large growth over the period of 1980 to 2006, with an annual growth rate of 0.46%. The average cash-to-assets ratios of US firms more than doubled from 10.5% in 1980 to 23.2% in 2006. They further figure out that the increase in cash ratios can be attributed to riskier cash flows of firms and the changes in firm characteristics. Their findings are confirmed by Azar, Kagy, and Schmalz (2014) who also point out that the current cash holdings of US firms are not abnormal. Another interesting finding of Bates et.al (2009) is that the overall increase in average cash-to-assets ratios cannot be explained by the evolution of cash holdings for large firms. They separate all the firms into different samples dependent on firm size, and they conclude that the increase in cash ratios exists for all size quintiles, not only for the relatively large firms.
were before the crisis, and although there is a slight increase in cash holdings, it is only concentrated among highly profitable large firms. This trend does not exist for other medium or small size firms.
The distinction in describing the evolution of cash holdings of US firms during the crisis raised my interests of how US firms actually have been affected by the global financial crisis with the focus on cash holding levels, and how firms responded to the global financial crisis by changing their cash holdings, and whether or not firms actually improved their cash positions after experiencing the crisis.
2.3 The Marginal Value of Cash Holding
In perfect capital markets, the wedge between the cost of internal and external funds does not exist, which means the value of one dollar in cash holding raised by the firm would be valued by shareholders with exactly one dollar and an extra one dollar held by the firm should also lead to an increase in firm value by exactly one dollar (Pinkowitz and Williamson., 2004). Nevertheless, with the concerns about imperfections of capital markets, a massive amount of literature depicts that the value of one additional dollar of cash rarely has the exact value of one dollar, and that the value of this one additional dollar of cash to shareholders should vary depending on the firm characteristics and shareholders’ expectation of how the additional one dollar of cash would be used.
value rather than the firm’s equity value so that the marginal value of one extra dollar cash also seems to be lower than $1 for shareholders. Whereas, only when the firms’ cash reserve is located in third regime that firms preserve the cash holding for the near future opportunities which decrease their need to outside capital markets, the value of one additional cash is supposed to be greater than $1. Faulkender and Wang (2006) regress the excess return1 on changes in firms’ cash holdings, control variables which reflect other firm characteristics and the interaction terms which measure the impact of increasing levels of cash and leverage on changes in the value of cash holdings. The control variables are used to control the effect of firm’s profitability, financing and investment policy on firm value. In order to correct for the influences of large firms, they apply the technique of deflating the firm characteristics as well as the previous cash levels except for leverage by the previous market value of equity, and Faulkender and Wang (2006, pp 1968) state that, “this standardization allows us to interpret the coefficients as the dollar changes in value to shareholders for a one-dollar change in the corresponding independent variables.” As a result, they find that the marginal value of cash holdings decreases when firms holding more cash and more leverage which is consistent with their expectations.
Other previous literature also indicates that the marginal value of cash holdings is a decreasing function of the improvement in levels of cash holding and leverage (Denis and Sibilkov, 2007 and Zhang, Von Eije and Westerman, 2015). Shareholders generally assess the additional cash less valuable when firms hold sufficient cash to face the short-term needs, but still raise their cash positions. The excess cash is more likely to be distributed to shareholders, but the ultimate value to shareholders would be discounted due to the taxes. Moreover, the agency theory appears to provide another proper explanation of why shareholders expect a lower value of cash since the excess cash held by firms increases the probability that the managers would pursue their own goals and invest the excess cash in value-decreasing projects. Simultaneously, shareholders would also regard an additional dollar of cash with a lower value for relatively high leveraged firms since the additional dollar of cash in
1 Faulkender and Wang (2006) constructed the excess (abnormal) stock return as dependent variable by deducting the
highly-leveraged firms would disproportionately increase the debt value of firms rather than equity value. Similar results have been confirmed by Zhang et al. (2015) who alter the regression model of Faulkender and Wang (2006) by treating the size and value effect as control variables.
2.4 Global Economic Shocks and the Subsequent Effect
Prior literature shows that macroeconomic conditions play an important role in determining the firms’ cash holding behavior. Song and Lee (2012) investigate the long-term effect of the Asian crisis of 1997-1998, and they find that Asian firms raised their cash positions by reducing their investment activities after the crisis. They further find out the increase in cash holdings is not attributed to the changes in firm characteristics but are a result of changes in the firms’ demand function for cash. This finding provides a strong evidence that the crisis has fundamentally and systematically changed firm’s cash holding policies.
The recent global financial crisis substantially destroyed the global financial conditions, and therefore, most US firms are inevitably affected by the macroeconomic shock. Firms’ cash flows became more risky and volatile during the crisis, and banks or other credit suppliers significantly limited their lending as a response to the crisis. McLean (2011) and Dittmar and Dittmar (2008) suggest that business cycles drive firms’ business activities, and generally, firms issue less debt, less shares and repurchase less in recessions and become more
active during expansions.Duchin et al. (2010) and Campello et al. (2010) report that the firms
formulated as follows:
Hypothesis 1: The crisis effect has a positive effect on the marginal value of cash holdings.
Hardly any literature has explored the post crisis effect on firms’ performance. Generally speaking, the business cycle theory underlines that firms invest less when markets decline and expand when markets bloom. Campello et al. (2010) indicate that firms prefer to postpone their planned investment until the crisis is over. The firms which survived from the crisis should hold more cash, at least for a short-term due to the precautionary motive (Song and Lee, 2012 and Kahle and Stulz, 2013). However, as a consequence of the increase in cash holdings, shareholders could become anxious about overinvestment. The managers with available internal funds could overinvest at the expense of shareholders’ value. Therefore, during the post crisis period, the marginal value of cash holdings might have a different features compared with the crisis period. As agency theory suggests, my second hypothesis can be formulated as follows:
Hypothesis 2: The post-crisis effect has a negative effect on the marginal value of cash holdings.
2.5 The Impacts of Financial Constraints
underinvestment problem (Opler et al., 1999). However, Pinkowitz and Williamson (2007) point out that the marginal value of cash is insensitive to the firms’ accessibility to external markets. Their findings are inconsistent with the majority of previous literature. This difference may be caused by the different methods to identify whether firms are constrained or unconstrained.
With respect to the classification of firms’ financial constraint, Faulkender and Wang (2006) provide four alternative schemes to define financially constrained firms. They use payout ratio, firm size, long-term bond rating and short-term commercial paper rating as the criteria to assign firms into constrained and unconstrained groups. In addition, Denis and Sibilkov (2007) propose additional two measurements and one of them is the KZ index, which is based on the contribution of Kaplan and Zingales (1997)2.
Theoretically, destructive shocks to credit supply markets, along with the presence of financing frictions, might discourage investment if firms have insufficient cash to fund all profitable investment opportunities internally (Duchin et al., 2010). Moreover, it suggests that such negative effects should be particularly severe for firms with greater difficulties to external capital markets. This means that the financially constrained firms are supposed to be more significantly affected by the financial crisis. Campello et al. (2010) provide evidence on the basis of their survey data from more than thousands CFOs, and demonstrate that financially constrained firms, generally, experienced more influence of the financial crisis by reducing more employment, capital investment and marketing expenditures.
On the other hand, the counter-argument exists, and it expresses that financially unconstrained firms might be affected by the crisis more heavily. Recall that during the crisis, banks, as well as other credit suppliers, significantly restricted their lending policies. The financially unconstrained firms depend more on credit prior to the crisis and the constrained firms are more likely to be the bank-independent firms (Chang, et al., 2016). According to the credit supply theory, Kahle and Stulz (2013), Campello, Giambona, Graham and Harvey (2012) state that firms that depend on external credit will be affected more by the crisis, and firms that rely less on credit should be affected less by the crisis. Additionally, the
2 Denis and Sibilkov (2007) report the KZ index is the weakest financial constraint classification compared to other
credit-dependent firms might be required to use their cash holdings to pay off their debt in order to reduce their default risk (Kahle and Stulz, 2013), and for those firms that do not need to renew debt have more flexibility to reduce capital expenditures to adjust to changes in the markets (Denis, 2011). From this point of view, the financially unconstrained firms might benefit more from the cash holdings which allows them to meet their labilities and to reduce the financial distress costs.
Hypothesis 3: The constraint effect has a constant and positive effect on the marginal value of cash holdings.
3. Empirical Methodology
According to the explanation of Faulkender and Wang (2006), the marginal value of cash holdings refers to the additional value shareholders obtain on an extra dollar of cash held by firms. They postulate that the marginal value of cash holdings depends on to which cash regime the firm is likely to belong. Furthermore, the accessibility of firms to external capital markets should also have an impact on the marginal value of cash holdings. I expand the regression model on the marginal value of cash holdings which is provided by prior literature by taking both the crisis effect and constraint effects into consideration. Chang et al. (2016) create a triple dummy to ascertain the different values of cash holdings 1) to the time period before and during the crisis and 2) to the external financing ability of firms. I further subdivide the time into three periods which are pre-crisis, during-crisis and post-crisis rather than two periods because I expect that the firms’ cash position and the motivation for holdings cash could be different after experiencing the global financial crisis, and that these differences could also exist between when firms are suffering from the crisis (during-crisis) and when firms have recovered from the crisis (post-crisis). This measure of time allows me to explore the combined effect of time and constraint more particularly on the value of cash holdings.
3.1 Evolution of Cash Holdings
Before testing the crisis effect and constraint effect on the marginal value of cash holdings, I first explore how the cash position of firms evolved over the period of 2002 to 2015. I have divided the time into the pre-crisis period (2002-2007), during-crisis period
(2008-2009)3 and post-crisis period (2010-2015) on the basis of the real GDP growth rate of
the United States4. The reason behind this is that the real GDP growth rate is identified as a
good indicator in economics and finance to describe the country’s economic environment.
3 There is no consensus on the timing of when the financial crisis started. Several studies indicate that bank borrowing was
sharply restricted from October, 2008 after the collapse of Lehman Brother (Ivashina and Scharfstein, 2010), while other studies point out that the crisis started already in the end of 2007 (Kahle and Stulz, 2013). Because in this study, annual data is used for the regression model, it seems to be illogical to include a couple of months of 2007 into the crisis period.
Figure 1 Real GDP Growth rate of the United States and OECD countries
Figure 1 demonstrates that the United States has experienced an apparent business cycle after 2000. It plunged after peaking in 2004, and reached the bottom in 2009. The real GDP growth rate sharply decreased in 2008-2009 with an obvious divergence from the average level, and it backed on track in 2010. The real GDP growth rate of OECD countries confirms the global economics behaved as the United States at each time period. By measuring the annual cash-to-asset ratio, I expect that the level of cash holdings should decrease in the crisis period due to the meltdown of global economy, it should tend to increase due to precautionary purpose after experiencing the global crisis.
3.2 Designed Regression Model
The baseline regression model refers to prior literature is: ri,t= β0+ β1∗ ∆Ci,t MVi,t−1+ β2∗ ∆Ei,t MVi,t−1+ β3∗ ∆NAi,t MVi,t−1+ β4∗ ∆Ii,t MVi,t−1+ β5∗ ∆Di,t MVi,t−1+ β6∗ Ci,t−1 MVi,t−1+ β7∗ Li,t+ β8∗ Ci,t−1 MVi,t−1∗ ∆Ci,t MVi,t−1+ β9∗ Li,t∗ ∆Ci,t Mi,t−1+ β10∗ MVTBi,t−1+ β11∗
SIZEi,t−1+ β12∗ Rmrfi,t+ εi,t, (Eq1)
where ∆Xi,t is defined as 1-year change in the variable X , accounts for the value of X for
3.2.1 Variable Definition
The dependent variable ri,t represents the stock return for firm 𝑖 during the year 𝑡 and
it is the changes in firm’s market value of equity divided by market value of equity of last year (MVi,t−MVi,t−1
MVi,t−1 ). This is inconsistent with Faulkender and Wang (2006) who construct the
excess return by deducting the benchmark portfolio returns from the stock raw returns as their dependent variable. In this study, I control for the market risk premium, size effect and value effect as independent variables instead of constructing the benchmark portfolios. This method is consistent with Zhang et al. (2015). I further regress the stock return on the changes in firm characteristics, with the most attention on cash holdings. The changes in cash holdings for firm 𝑖 during the year 𝑡 are expressed as ∆Ci,t. I follow the technique applied by the prior literature (Faulkender and Wang, 2006) to deflate changes in cash by dividing it by the previous year market value of equity MVi,t−1.5 The estimated coefficient of changes in cash holdings is anticipated to be positive because of the fact that the firms with sufficient cash can afford future profitable investments which can enhance the firm value, and this is reflected in the share price. To take into account the existence of asymmetric information and reporting lag, stock markets do not always immediately respond to the changes in firm characteristics, I
also regress the returns of subsequent year 𝑡 + 1 as dependent variable on current changes in
firm characteristics during the year 𝑡 except for those variables may not suffer from a lag. The rationale in applying the next year stock return as dependent variable lies in the fact that there exists asymmetric information between the shareholders and managers and a reporting
lag of firms’ performance. Shareholdersoccasionally react slowly to new information so that
the equity value of firms is also not being affected immediately.
3.2.2 Control Variables
As we are interested in the relationship between the equity performance and changes in cash holdings, it is important to control for other firm specific characteristics that have an influence on firm value. Following previous literature (Faulkender and Wang, 2006 and Zhang et al., 2015), several control variables are included in the regression model6: Ei,t
measures the earnings before interest and extraordinary items and 𝑁𝐴𝑖,𝑡 indicates the net
asset of the firm; 𝐼𝑖,𝑡, 𝐷𝑖,𝑡and 𝐿𝑖,𝑡 indicate the interest expense, total dividend payment and
market leverage respectively. Furthermore, as mentioned before, market risk premium, 𝑅𝑚𝑟𝑓𝑖,𝑡 is also controlled as independent variable because I believe that the market return also has an impact on individual return. Firm size and market-to-book value are incorporated as control variables in the regression model by lagged value (𝑆𝐼𝑍𝐸𝑖,𝑡−1 and 𝑀𝑉𝑇𝐵𝑖,𝑡−1) at the beginning of year 𝑡. The control variables applied in this study are consistent with Zhang et al. (2015) except for the market risk premium. In addition, all the control variables except for leverage, firm size and market-to-book ratio, are scaled as changes in cash holdings by deflating the changes in each variables by the lagged year market value of equity 𝑀𝑉𝑖,𝑡−1.
Several control variables applied by previous studies are excluded in this study because of the unavailability of data sources, for example R&D expenses and net financing. The inclusion of the interaction terms, Ci,t−1
MVi,t−1∗ ∆Ci,t
MVi,t−1 and Li,t∗ ∆Ci,t
MVi,t−1 allows me to investigate the effect of
changes in the value of cash for different levels of cash holdings and different levels of leverage. In order to ensure that the effect of the interaction terms is estimated appropriately, I also add the lagged level of cash holdings into the regression.
In the case of regressing the next year stock return (𝑡 + 1) as dependent variable, most independent variables maintain the current level of year 𝑡, while dividend and market risk premium are correspondingly changed to the next year (𝑡 + 1) due to the facts that the dividend payout should be immediately known by the shareholders, so that the dividend effect on share price should be instant, and the market effect on individual stock is supposed to be instant as well. The variables control for the size effect (𝑆𝐼𝑍𝐸𝑖,𝑡−1) and value effect (𝑀𝑉𝑇𝐵𝑖,𝑡−1) in this case is adjusted to the current year level, and becomes 𝑆𝐼𝑍𝐸𝑖,𝑡 and
𝑀𝑉𝑇𝐵𝑖,𝑡 which is now consistent with other firm characteristics variables.
3.3 Time Period Specifications
ri,t= β0+ β1∗ ∆Ci,t MVi,t−1+ β2∗ ∆Ei,t MVi,t−1+ β3∗ ∆NAi,t MVi,t−1+ β4∗ ∆Ii,t MVi,t−1+ β5∗ ∆Di,t MVi,t−1+ β6∗ Ci,t−1 MVi,t−1+ β7∗ Li,t+ β8∗ Ci,t−1 MVi,t−1∗ ∆Ci,t MVi,t−1+ β9∗ Li,t∗ ∆Ci,t MVi,t−1+ β10∗ MVTBi,t−1+ β11∗
SIZEi,t−1+ β12∗ Rmrfi,t+ β13∗ Ddurcisis𝑖,𝑡∗ ∆Ci,t MVi,t−1+ β14∗ Dpostcrisis𝑖,𝑡∗ ∆Ci,t MVi,t−1+ β15∗ Ddurcisis𝑖,𝑡∗ Ci,t−1
MVi,t−1+ β16∗ Ddurcisis𝑖,𝑡 + β17∗ Dpostcrisis𝑖,𝑡+ εi,t, (Eq2)
where Ddurcrisis is the time dummy variable for the time period during the crisis, equals 1
for years of 2008 and 2009 and equals 0 the rest of years; Dpostcrisis represents the time period after the crisis (post- crisis), equals 1 for the period from 2010 to 2015 and equals 0 for the years before 2010. I also include the period dummy as a separate variable to ensure the accuracy of coefficients. The interaction-term of Ddurcisis𝑖,𝑡∗ Ci,t−1
MVi,t−1 is added in the
regression model with the consideration of that the cash level at the start of the crisis also
have an impact on firms’ equity performance. The major estimates are β13 and β14 which
reflect the time effect on the marginal value of cash holdings. In this case, β13 is expected to be positive since the global financial crisis destroys firm’s performance during the crisis period, any additional one dollar cash should have a higher marginal value than before. Reversely, β14 is expected to be negative. This is in line with the expectation of my first research question that firms would increase their level of cash holdings after the crisis. The increasing level of cash holdings results in a lower marginal value of cash holdings. A further regression is conducted as explained in section 3.2.1 with the modifications of variables are explained in section 3.2.2. The asymmetric information and reporting lag exist in an imperfect market, shareholders always suffer from an information lag of firms’ performance so that I also regress the next year stock return as dependent variable. Meanwhile, dividend and market risk premium, as variables may not suffer from the reporting lag, and is correspondingly modified to the next year change level which is consistent with the stock return.
3.4 Classification of Financial Constraints
value of cash holdings, I construct another dummy variable DFC𝑖,𝑡 that equals 1 for the firms
being regarded as financially constrained firms during the year 𝑡 and equals 0 for the firms
that are financially unconstrained. Two triple-interaction terms are created to measure the combined effect of crisis (post-crisis) effect and constraint effect. The financial constraint dummies are also added separately to control for any differences in the valuation of constrained and unconstrained firms (Denis and Sibilkov, 2010).
The regression model for third research question therefore turn into as follow: ri,t= β0+ β1∗ ∆Ci,t MVi,t−1+ β2∗ ∆Ei,t MVi,t−1+ β3∗ ∆NAi,t MVi,t−1+ β4∗ ∆Ii,t MVi,t−1+ β5∗ ∆Di,t MVi,t−1+ β6∗ Ci,t−1 MVi,t−1+ β7∗ Li,t+ β8∗ Ci,t−1 MVi,t−1∗ ∆Ci,t MVi,t−1+ β9∗ Li,t∗ ∆Ci,t MVi,t−1+ β10∗ MVTBi,t−1+ β11∗
SIZEi,t−1+ β12∗ Rmrfi,t+ β13∗ Ddurcisis𝑖,𝑡∗ ∆Ci,t
MVi,t−1+ β14∗ Dpostcrisis𝑖,𝑡∗ ∆Ci,t MVi,t−1+ β15∗ Ddurcisis𝑖,𝑡∗ Ci,t−1 MVi,t−1+ β16∗ DFC𝑖,𝑡∗ ∆Ci,t MVi,t−1+ β17∗ DFC𝑖,𝑡 ∗ Ddurcisis𝑖,𝑡∗ ∆Ci,t MVi,t−1+ β18∗ DFC𝑖,𝑡∗ Dpostcrisis𝑖,𝑡∗ ∆Ci,t
MVi,t−1+ β19∗ Ddurcisis𝑖,𝑡+ β20∗ Dpostcrisis𝑖,𝑡+β21∗ DFC𝑖,𝑡 +
β22∗ DFC𝑖,𝑡∗ Ddurcisis𝑖,𝑡 + β23∗ DFC𝑖,𝑡∗ Dpostcrisis𝑖,𝑡+ εi,t , (Eq3)
The limited access to external capital markets should place a positive effect on the valuation of the extra dollar held by the firms as cash. In other words, the financially constrained firms should have a higher marginal value of cash holdings which means β16
should be positive. In addition, the constraint effect on value of extra cash is expected to be constant over time. This expectation gives rise to that β17 and β18 should be all positive,
and it means financially constrained firms should always have a higher marginal value of cash no matter during or after the crisis.
In order to construct the financial constraint dummy, three alternative approaches are used in this study according to the availability of information. All these three approaches are extensively used by a great number of articles on classifying financial constraints.
decile are assigned to the financially constrained (unconstrained) group, and those firms located in four deciles are left out.
2. Debt rating (long-term bond rating): A majority of articles propose that the rating index can be regarded as a good indicator of financial constraint of firms. Whited (1992), Almeida et al. (2004), Faulkender and Wang (2006), Denis and Sibilkov (2010) state that the firms are classified as financially unconstrained if they have a long-term debt rating provided by Standard & Poor’s and no records of default, because these firms are more likely to be well known, and should have less difficulty in raising capital. In this study, I follow the same criterion as adopted by Faulkender and Wang (2006). That is, firms are assigned to the financially constrained group if they have no available long-term debt rating on COMPUSTAT, but have positive debt that year, or they have a default record in debt that year. Conversely, firms with available long-term debt rating or no debt outstanding are classified as financially unconstrained that year.
3. Paper rating (short-term commercial paper rating): Similar to debt rating, paper rating focusses on short-term commercial paper rating. Firms with unavailable short-term debt rating on COMPUSTAT but positive debt at that year are classified as financially constrained firms.
4. Data and Summary Statistics
4.1 Sample Selection
In this article, two panel data samples are constructed of United States public firms during the period from 2002 to 2015 with available firm-level information in Obris and Datastream (Worldscope database). The firms with unavailable ISIN code and Ticker Symbol are eliminated from the sample.7 Financial firms and utility firms with SIC code between 6000-6999 and 4900-4999 are excluded from the sample due to the possibility that these firm may have different reasons to hold cash. Financial firms are inclined to hold cash or securities as inventory, which is not related to their performance but just because of the regulation on minimum capital requirements, while utility firms are subject to regulatory supervision in a number of states (Opler et al. 1999). Another restriction is that only surviving firms are
7 ISIN codes are required for Datastream to find firms’ market and account data, and Ticker Symbols are required for
included in the sample given that non-surviving firms were struggling at some specific time period during the whole sample period. Therefore, adding non-surviving firms into samples
might distort the whole cash holding trends.8 Delisted firms are also obviated for the reasons
of missing available stock returns. In fact, the only four firms are precluded from the sample as result of delisting, hence, the effect of not including delisted firm should be minor.
The majority of variables in the regression are sourced from Datastream (Worldscope database). Cash holdings refer to cash and cash equivalent. Earnings are constructed by net income before extraordinary items plus interest expenses, and net assets are constructed by total asset minus cash holdings. Interest expenses are directly sourced from Datastream and dividend is defined as total cash common dividend. Leverage is measured as total debt divided by the sum of market value of equity and total debt during the same fiscal year. Market to book ratio is provided by Datastream and firm size is measured as natural logarithm of total assets. The market risk premium is sourced from the Fama-French online database, which is constructed as market return minus risk free rate and the financial constraints classification of debt and paper rating are collected from COMPUSTAT based on Ticker Symbol.9 Following Faulkender and Wang (2006), Zhang et al. (2015) and Chang et al. (2016), the firm-year observations with negative net assets and dividend are excluded.
Recall that most independent variables except for Leverage 𝐿𝑖,𝑡 , market risk
premium 𝑅𝑚𝑟𝑓𝑖,𝑡, firm size 𝑆𝐼𝑍𝐸𝑖,𝑡and market to book value 𝑀𝑉𝑇𝐵𝑖,𝑡 are constructed in the form of dividing by lagged market value of equity 𝑀𝑉𝑖,𝑡−1. Simultaneously, because several
variables are estimated at change level, the data of one prior year (2001) is required for building the entire sample from 2002 to 2015. In addition, I winsorized both dependent and independent variables at 1% and 99% tails in order to preclude the outlier effects. The whole sample then consists of 22445 firm-year observations according to those sample selection criteria, and additionally, a sub-sample with 1163 firms (16338 firm-year observations) that has complete data during the whole period of 2002-2015 are created for the purpose of investigating the sensitivity of coefficients between the balanced panel data (sub-sample) and
8 The restriction of excluding the non-surviving firms also has a limitation that it could lead to survivor bias.
9 The debt rating and paper rating refer to the long-term and short-term debt rating respectively, and the data are provided
the unbalanced panel data (whole sample).
4.2 Descriptive Statistics
Table 1 reports the descriptive statistics of the variables for the balanced panel sample (sub-sample) and the unbalanced panel sample (whole sample). The most compelling finding is that the raw stock return as the dependent variable which measured differently compared to Faulkender and Wang (2006) but shows a quite similar characteristic. Faulkender and Wang (2006) report that their abnormal stock returns are skewed to the right as the median firm has a much lower abnormal stock return than the mean abnormal return of the whole sample10.
This is consistent with the statistics of stock return in this study, and the descriptive statistics indicate that the mean stock return is much higher than the median stock return for both samples, 0.264 and 0.103 for the balanced panel sample, meanwhile, 0.344 and 0.088 for the unbalanced panel sample. Moreover, this distribution of stock return also shows a similarity to Zhang et al. (2015) who have same measurement of the raw stock return as dependent variable.
With respect to the dependent variables, the two samples share similar features. Table 1 reports that the changes in cash holdings are generally increasing during the period because of the slightly positive mean and median. The changes in interest expense and dividend payment fluctuate around zero which is consistent with previous studies. Another similarity is that on average, profitability shows an upward tendency during the last 15 years because of the positive changes in earnings. However, market to book value appears to be a bit strange variable owing to the negative market to book ratios existing in both samples. Because of the limited liability of shareholders, it means that both samples contain firms with negative book value of equity. The significance of correlations between variables are tested by T-test and are exhibited in Table 2. The correlation coefficients are significantly low for most cases, and the highest correlation occurs between the lagged cash holdings and stock return with 0.335 for the sub-sample (0.341 for the whole sample).
In order to test the constraint effect on the marginal value of cash holdings, the financial
Table 1 Summary Statistics of Samples for the period of 2002-2015
This table presents the descriptive statistics which include mean, median, standard deviation, maximum value, minimum value and observations of main variables for two sample-sets during the period of 2002-2015. Panel A shows the statistics of the sub-sample (balanced panel, 1163 firm-14 years) that winsorized at 1% and 99% tails, while Panel B shows the statistics of the whole sample (unbalanced panel, 1714 firms-14 years) that winsorized at 1% and 99% tails. 𝑟𝑖,𝑡 is the stock return for the firm 𝑖 during fiscal year 𝑡. 𝐶𝑡 is cash and cash equivalent; 𝐸𝑡 is earnings which is constructed by net income before extraordinary items plus interest expense; 𝑁𝐴𝑡 represents the net assets, constructed by total asset minus cash holdings; 𝐼𝑡 is interest expense and 𝐷𝑡 is total cash common dividend; 𝐿𝑡 account for leverage which is measured as total debt divided by the sum of market value of equity and total debt during the same fiscal year; 𝑀𝑉𝑇𝐵𝑡−1 is market to book ratio and 𝑆𝐼𝑍𝐸𝑡−1 is measured as natural logarithm of total assets (both these two factors are controlled as size effect and value effect). 𝑅𝑚𝑟𝑓𝑖,𝑡 represents for the market risk premium. ∆𝑋𝑖,𝑡 is compact notation for 1-year change, 𝑋𝑖,𝑡− 𝑋𝑖,𝑡−1. All the variable except for 𝑟𝑖,𝑡, 𝐿𝑖,𝑡, 𝑀𝑉𝑇𝐵𝑖,𝑡−1 and 𝑆𝐼𝑍𝐸𝑖,𝑡−1 are divided by the lagged market value of equity (𝑀𝑉𝑡−1).
Variable N. Mean Median Std.dev Max. Min. N. Mean Median Std.dev Max. Min.
Table 2 Correlation Matrix
This table contains the correlations coefficients between dependent variables and independent variables. All variables have the same definition as explained above. Panel A represents for the sub-sample with complete data and Panel B represents for the whole sample. ***, **, * represent for significance level 1%, 5%, and 10%.
Table 3 Correlation of Different Classification of Financial Constraints
The table indicates the correlation between various classifications of financial constraints. Panel A represents for the sub-sample with complete data and Panel B represents for the whole sample. The correlation coefficients reported with the p-value in parentheses.
Financial Constraints Criteria
Firm Size Debt Rating Paper Rating
Panel A Firm Size 1 Debt Rating 0.6425 (0.0000) 1 Paper Rating 0.1904 (0.0000) 0.6249 (0.0000) 1 Panel B Firm Size 1 Debt Rating 0.5177 (0.0000) 1 Paper Rating 0.0433 (0.0000) 0.5904 (0.0000) 1
constraint dummy is included in the regression model. Three financial constraint classification criteria have been introduced: two of them are directly sourced from the COMPUSTAT database and the other one is measured by ranking firm size. Table 3 shows the correlations between alternative measures of financial constraints. All the correlation coefficients are strongly significant and positively correlated within the range of 0.043 to 0.643. Only the coefficient between firm size and paper rating in the whole sample with incomplete data (in Panel B) is considerably lower than other correlation coefficients. Denis and Sibilkov (2010) concluded that although the high correlations raise the concerns about the similarity of information between measurements, it nevertheless appears that each measurement provides some unique information as well.
5. Empirical Results and Discussion
5.1 Stationarity
hypothesis of existence of a unit root are rejected statistically significant (at 1% significant level) for all variables11. This means that the stationarity condition is satisfied and the variables can be employed for the following regressions.
5.2 Results of Cash Holdings Dynamics
Before regressing equity performance of firms on changes in cash holdings, I first explore the dynamics of cash holdings over the sample period from 2002 to 2015. I expect that firms should have relatively lower cash holdings during the crisis (2008-2009) due to the recession of the global economy, and then they should increase their cash holdings after the crisis (2010-2015), at least in the short-run. Table 4 reports the annual cash-to-asset ratios for the sub-sample firms over the period. The result is consistent with the expectation as the mean cash-to-asset ratio decreased in 2008, and peaked at 19.65% in 2010. The changes of median cash-to-asset ratio confirms the overall trends of how firms changed their cash positions over the years. If we focus on the changes in cash holdings over the different periods, Table 4 (Panel B) shows relatively weak evidence that on average, during the crisis-period, firms have the lowest mean cash-to-asset ratio of 18.11%, and firms have the highest mean cash-to-asset ratio of 18.58% in the post-crisis period. In general, the mean (median) cash-to-asset ratios show much stronger evidence on an annual basis (Panel A). Figure 2 graphically exhibits the dynamics of the cash holdings of US firms. Consistent with the results of Table 4, it is apparent that the cash/asset ratio decreased to the bottom both at the mean and the median during the crisis, and it climbed up immediately after the crisis and maintained a stable level in recent years. The cash-to-assets ratios for the whole sample show a similar pattern as the sub-sample show12.
The reasons why firms held less cash during the financial crisis are quite straightforward. The global financial crisis severely exacerbated firms’ performance. Either regular operation income or investment returns of firm are dramatically suppressed as a consequence of the recession. It is conceivable that a large fraction of firms experienced a low cash flow during this period, and therefore they held relatively lower cash on hands. However, there could be
11 The test coefficients are significantly higher than any critical value, and all p-values are equal to 0.
Table 4 Cash-to-Assets Ratios for the Sub-sample Firms during the period 2002-2015 This table provides the cash to assets ratios for the sample firms. Mean Cash/Asset represents the average cash to assets ratio of all firms, and Median Cash/Asset is the median cash to assets ratio of all firm. Panel A shows the ratios on the year basis and Panel B shows the ratios on the corresponding sub-period basis.
Year N Mean Cash/Asset Median Cash/Asset
Panel A 2002 1163 16.43% 7.48% 2003 1163 18.21% 9.55% 2004 1163 19.16% 10.33% 2005 1163 19.18% 10.50% 2006 1163 19.16% 10.26% 2007 1163 18.12% 9.61% 2008 1163 16.77% 8.98% 2009 1163 19.44% 12.21% 2010 1163 19.65% 13.09% 2011 1163 19.01% 11.82% 2012 1163 18.29% 11.01% 2013 1163 18.49% 11.67% 2014 1163 18.27% 11.19% 2015 1163 17.74% 10.39% Panel B Pre-Crisis 6978 18.37% 9.62% Dur-Crisis 2326 18.11% 10.60% Post-Crisis 6978 18.58% 11.61%
various motivations why firms have raised their cash holdings after the financial crisis. On one hand, firms might have increased their cash holdings after experiencing the financial crisis owing to the precautionary purpose. Firms who deeply suffered from the crisis are inclined to increase their cash positions to prevent themselves from getting to any forms of financial distress. On the other hand, firms’ cash position could increase simply because firms invested less (Pinkowitz et al., 2013). Due to the poor regulatory and financial climate, firms generally became more conservative in investing in projects with high uncertainty. As a result, they preferred holding cash for future projects with comparably higher certainty instead of bearing risks. These expositions explain the increasing trends in cash positions of firms after the crisis.
Figure 2 Cash/Assets Ratios of the Sub-sample Firms During the Period 2002-2015
pre-crisis period starting from 2003. Second, I find that even though the US economy suffered from the impact of the financial crisis during the years 2008-2009, the cash-to-asset ratio indicates that the firms may only have suffered from a cash shortage in 2008, and that they replied quickly to the crisis by improving their relative cash holdings in 2009. According to Figure 2, I construct an alternative crisis dummy which reflects that the crisis period only includes 2008, and correspondingly, the post-crisis dummy starts from 2009 to 2015.
5.3 The Effect of Crisis (Post-crisis) on Cash Holdings
The results of the crisis effect and the post-crisis effect on the marginal value of cash holdings for the sub-sample are presented in the Table 5. The regressions are conducted based on Eq2, and simultaneously controlled for industry fixed effect. By regressing the contemporaneous stock returns (stock return at year t) as dependent variable, as Panel A shows most of the firm-specific characteristics have the expected association with stock returns and are statistically significant. The market risk premium has an opposite sign against my expectation, and it might be the case that the market effect has been captured by the crisis and post-crisis dummies themselves. The significantly positive estimated coefficients of change in cash holdings ( β1 ) in both columns confirms the cash effect on enhancing firm value. However, if we focus on column 2, in which the crisis period consists of the years 2008 and 2009, and the post-crisis period starts from 2010, the crisis effect and the post-crisis effect go against my expectations. The estimate of crisis effect ( β13 ) turns out to be negative
instead of the expected positive sign, while the estimate of the post-crisis effect ( β14 ) turns out to be insignificant instead of being negative. β1, the marginal value of one extra cash
Table 5 The Crisis Effect and Post-crisis Effect on the Marginal Value of Cash Holdings for the Sub-sample
This table shows the results of regressing stock return 𝑟𝑖,𝑡 on 1) changes in firm characteristics over the fiscal year, on 2) interaction-terms of multiplying lagged cash holding and leverage on cash holdings and 3) interaction-terms of multiplying period dummies on cash holdings. Durcrisis (Postcrisis) are period dummies which have the meaning of 1 for crisis period (post-crisis period). Other firm characteristics variable have the same definition as mentioned before. Panel A reports the outcomes for applying the contemporaneous stock
returns as dependent variable, while Panel B reports the outcomes for applying the next year stock returns as
dependent variable. The results for during the crisis period 2008-2009 are reported in column 2 and column 4, and the results for during the crisis period 2008 are reported in column 3 and column 5. Coefficients are reported with the standard error in parentheses. ***, **, * represent for significance level 1%, 5%, and 10%.
for the firms before the global financial crisis (pre-crisis period) is 1.605, and during the crisis period, the marginal value in equity market of an extra one dollar change in cash decreases to 1.261 (1.605 - 0.344). The insignificance of the post-crisis effect manifests that the marginal value of cash holdings of firms in the pre-crisis period and the post-crisis period seems to share the same patterns. The overall patterns of effects of firm characteristics on stock return are similar when the alternative crisis dummy (only 2008) is applied. The crisis effect on the marginal value of cash holdings in this case becomes insignificant and the post-crisis effect is also insignificant.
Another interesting finding is that the coefficients of interaction terms of the changes in cash with the level of cash holdings and the level of leverage ( β8 and β9 ) are significantly negative in both cases which confirms the diminishing function of marginal value of cash holdings to cash positions and leverage (Faulkender and Wang, 2006 and Denis and Sibilkov, 2007). Numerically, again if we focus on column 2, the coefficient of the interaction-term 𝐶𝑖,𝑡−1∗ ∆𝐶𝑖,𝑡 is -0.434 which can be interpreted as that for two identical firms, the marginal
value of cash holding for the firm with cash holding of 10% of equity would be approximately 0.434 higher than the firm with cash holding of 11% of equity. The higher the cash position, the lower the marginal value of cash. This relationship also holds for leverage. The coefficient of 𝐿𝑖,𝑡 ∗ ∆𝐶𝑖,𝑡 of -1.122 indicates that a difference in the leverage ratio of 1% would
contribute to a difference of 1.122 dollars in the marginal value of cash holdings. Faulkender and Wang (2006) explained that debt holders and equity holders jointly benefit from an increase in the amount of cash the firm holds, but the value of cash to equity holders is more likely to be higher when the firm has relatively lower levels of debt since the possibility of default is lower when the firm holds less debt, and the value of additional cash is more likely to accrue to equity holders. The coefficient of the interaction-term Ddurcisis𝑖,𝑡∗ 𝐶𝑖,𝑡−1 indicates that the cash level at the start or during the crisis has a negative impact on stock return.
Considering that the stock markets may react slowly to the changes in firm specific characteristics since there exists asymmetric information and a reporting lag on firms’ operating performance, I also regress the stock return of subsequent year 𝑡 + 1 on the
variables of changes in firm characteristics at year 𝑡 except for the dividend and market risk
returns. The interaction terms 𝐶𝑖,𝑡−1∗ ∆𝐶𝑖,𝑡 and 𝐿𝑖,𝑡∗ ∆𝐶𝑖,𝑡 still have a negative association with stock returns, but the effect of additional leverage on the marginal value of cash holdings decreases from -1.122 to -0.457. The interaction terms can still be explained by that shareholders would place a lower value on extra cash holdings when the firm increases the levels of cash holdings or leverage. The role of cash level during the crisis on firms’ stock return in the case of reporting lag is ambiguous. It has either a positive or negative effect based on different definitions of crisis period. The most striking finding of Panel B is that, taking column 4 as example, the crisis effect, β13 becomes positive in this case, and the post-crisis effect, β14 becomes negative in this case of the crisis period defined as 2008-2009. The marginal value of cash holdings before the crisis ( β1 ) reported is equal to 1.754 and
during the crisis, the marginal value of cash holding ( β1+ β13 ) increases to 2.127 (1.754 +
0.373), while, after the crisis, the marginal value of one additional dollar change in cash ( β1+ β14 ) drops to 1.377 (1.754 - 0.377). The results in case the crisis period that only include 2008 are inconsistent with the results of in case the crisis lasts for two years (2008-2009). The crisis effect on the marginal value of cash holdings is significantly negative (-0.342), while the post-crisis effect remains negative.
As explained before, I restricted both samples from 2003 to 2015 because of the finding that US firms were still experiencing recession in 2002. The regression with this restriction shows similar results, which ensures that including 2002 in the pre-crisis period would not have any influences on outcomes. Furthermore, due to the results that the market risk premium ( β12 ) unexpectedly has a negative relationship with stock returns and the crisis dummy, β16 (post-crisis dummy, β17 ) is significantly negative (insignificant). It might be
the case that the effect of the market premium has been partly captured by those period dummies. Thus, in order to examine if this is the case, I left out the market risk premium, and again the outcomes are consistent with the previous results. As a check, the results for the whole sample are reported in the appendix Table A3. The crisis effect and the post-crisis effect have a similar pattern between the sub-sample and the whole sample.
5.4 The Effect of Constraint on Cash Holdings
Table 6 & Table 7 report the regression results for the sub-sample based on Eq3, which analyses the effects of changes in cash holdings on firm value (equity market performance),
for constrained and unconstrained firms, before, during and after the global financial crisis.13
Specifically, Table 6 shows the result of regressing the contemporaneous stock returns as dependent variable but with alternative classifications of crisis period and financial constraints, while Table 7 reports the results of applying the next year stock returns as
dependent variable.14 To measure the difference in the value of cash between constrained and
unconstrained firms, three alternative financial constraints criteria are introduced to construct the financial constraint dummy: firm size refers to the measurement that firms are ranked by their annual total assets during each year, and the firms in the bottom (top) three deciles are assigned into financially constrained (unconstrained) groups that year; debt rating is based on the firms long term credit ratings, firms have unavailable debt rating but with positive debt, or any records of default in debt are assigned to the financially constrained group that year, otherwise they are classified into the unconstrained group; paper rating shares similar criteria as debt rating but with the emphasis on the short term commercial paper ratings. The interaction-term 𝐹𝐶 ∗ ∆𝐶𝑖,𝑡 estimates the constraint effect on the marginal value of cash
holdings, moreover 𝐷𝑢𝑟𝑐𝑟𝑖𝑠𝑖𝑠 ∗ 𝐹𝐶 ∗ ∆𝐶𝑖,𝑡 and 𝑃𝑜𝑠𝑡𝑐𝑟𝑖𝑠𝑖𝑠 ∗ 𝐹𝐶 ∗ ∆𝐶𝑖,𝑡 estimates the
combined effect of constraint effect and crisis (post-crisis) effect on the marginal value of extra cash held by firms.
5.4.1 Contemporaneous Constraint Effect on Cash Holdings
Table 6 shows main results of regressing the contemporaneous stock returns on the firm characteristics and the interaction-terms measured for crisis (post-crisis) effect and constraint effect. The changes in cash holdings ( β1 ) are all significantly positive, varying from 1.304 to 1.756 according to three alternative classifications of financial constraint and two alternative classifications of crisis period. It means that the excess cash holdings would enhance the firm value. Simultaneously, all control variables are in line with the expectations and the results of the previous research question regressions.
The interaction-terms between changes in cash holdings and the crisis period dummy as
well as post-crisis period dummy which have the estimates of β13 and β14, account for the
crisis and post-crisis effect for financially unconstrained firms. As a result, the crisis effect ( β13 ) is only significantly negative for one of the three classifications when the crisis dummy
is constructed of 2008-2009, and none of them turns out to be significant when the crisis dummy is constructed of 2008. The post-crisis effect is insignificant for every classifications no matter how the crisis dummy is defined. The outcomes show that for unconstrained firms,
14 The regression results for either the sub-sample or the whole sample based on each financially constrained criteria will
Table 6 The Contemporaneous Constraint Effect on the Marginal Value of Cash Holdings for the Sub-Sample
This table presents the main results of regressing stock return on 1) changes in firm characteristics over the fiscal year, 2) interaction-terms of multiplying lagged cash holding and leverage on cash holdings, 3) interaction-terms of multiplying period dummy on cash holdings, 4) interaction-terms of constraint dummy on cash holdings and 5) interaction-terms of both period dummy and constraint dummy on cash holdings. The financial constraint criteria based on firm size, long term debt rating and short term paper rating are applied for constructing constraint dummy which equals 1 for financially constrained firm and 0 for financially unconstrained firm. Column 2,4,6 with the top symbol I report the results of regarding 2008-2009 as crisis period, and column 3,5,7 with top symbol II report the results of regarding only 2008 as crisis period. Coefficients are reported with the standard error in parentheses. ***, **, * represent for significance level 1%, 5%, and 10%.
Firm Size Debt Rating Paper Rating
the marginal value of cash is equal for the post-crisis period and pre-crisis period. Meanwhile, the coefficients of the interaction term Ddurcisis𝑖,𝑡∗ 𝐶𝑖,𝑡−1 (β15 ) are significant and negative for five of six by considering the different classifications of constraint and crisis period. This is consistent with the results in section 5.3 which means cash level itself at the start or during the crisis negatively affects stock return during the crisis.
Unexpected outcomes occur for financially constrained firms. The interaction of changes
in cash holdings and the constraint dummy ( β16 in Eq3) turn out to have a very inconsistent
result between different classifications of financial constraint. Table 6 indicates the coefficient of β16 is positive in the case financial constraints are linked to firm size, while it is negative and insignificant in the case of debt rating and paper rating. It means that before the crisis, the constraint effect on marginal value of cash is ambiguous. This is inconsistent with prior studies (Faulkender and Wang, 2006, Denis and Sibilkov, 2007 and Chang et al., 2016). With respect to the crisis and post-crisis effects on the marginal value of cash holdings for constrained firms which are represented as β17 and β18 in Eq3, the results show that the
crisis effect and the post-crisis effect for constrained firms are generally insignificant in most cases, only with an exception for the post-crisis effect based on the classification of long term debt rating. It means the marginal value of cash holdings is very likely to be same for both financially constrained and unconstrained firms during the crisis and after the crisis. The shareholders show no difference in treating the extra dollar held by firms between financially constrained and unconstrained firms.
Considering that the size effect might be double-controlled by the variable of firm size and the constraint dummy based on firms’ total assets. As a check, I left out the size variable ( β11 ) when the financial constraints are linked to firm size. The regression results without the size variable are in line with the previous results except for the coefficient of the interaction of changes in cash holdings and the constraint dummy ( β16 ) which is insignificant instead of
