Cash holdings in Europe and the United States
The development before, during and after the financial crisis
Master thesis in International Financial Management Supervisor: Dr. W. Westerman
04.06.2020
Name: Marc-Oliver Polten
Student Number: S3759733
Program: MSc IFM
Key Words: Cash management;
I
Abstract
1
1. Introduction
Cash holdings of firms in the United States (U.S.) have been a highly discussed issue in the literature. After a strong increase of cash ratios from the 1980s to 2000s, the cash holding behavior of firms arose public attention. Negative effects of firms hoarding high amounts of cash, such as under-investment or low returns on cash, have been debated. Hence, the interest of research in the determinants of corporate cash holding behavior increased. Opler et al. (1999) hold that transaction cost considerations and information asymmetry theory best explain U.S. cash holding behavior. However, they also find some evidence for agency effects as well as the classical hierarchy theory. More recently, Bates et al. (2009) find that the reason for increased cash ratios in the U.S. are uncertainty avoidance and changing firm attributes such as higher R&D intensity and lower inventories.
2
only observable in the U.S. Hence, they propose that higher U.S. cash holdings are determined by firm-specific characteristics. Pinkowitz et al. (2013) implement a model of abnormal cash holdings which compares true cash holdings with those predicted by a model based on an earlier period. They find that since the late 1990s, U.S. abnormal cash holdings increased more than in other regions of the world. In the sample period 2001-2010, they specifically find that Eurozone countries have significantly lower abnormal cash holdings than U.S. firms. Iskandar-Datta and Jia (2012) analyze how cash holding patterns differ globally by conducting separate regressions for a group of seven industrialized countries from around the world. They find that cash practices indeed diverge and the increase of cash ratios can only partly be explained with changing firm attributes. Also, evidence for agency effects is found. However, with a strong focus of the literature on U.S. firms, the cash holding of European firms has been underexplored, especially in the time frame after the financial crisis. Figure 1 depicts a time plot of country averages comparing Western European and U.S. cash ratios. In the following, the term European refers to the considered country group of 17 Western European countries which are in scope of this paper.
3
Based on Figure 1, I can confirm previous findings related to the general increase in average cash ratios and indicate that European average cash holdings are lower but develop similarly over time in Europe and the U.S.
4
To answer my research questions, I use a large firm-level panel dataset for U.S. and European firms for 2002 to 2018. Confirming previous literature, I find that means and medians of U.S. firms’ cash ratios are significantly larger than those of European firms. Furthermore, differences between the various considered European countries appear to exist. Especially, the Anglo-Saxon influenced countries, including the United Kingdom (U.K.) and others, show significantly lower cash-to-assets ratios than Continental European countries. However, these findings could be due to differences in firm characteristics and hence, be not country-specific. To further analyze this issue, I deploy two main econometric approaches.
5
increased need for cash per unit of cash flow risk is a main determinant of higher cash ratios in the U.S. compared to Europe. Hence, in line with Pinkowitz et al. (2016), my findings suggest that firm differences are the major cause for higher U.S. cash ratios.
Second, I deploy a model of abnormal cash holding similarly to that of Pinkowitz et al. (2013). Based on this, I calculate excess cash holdings that are not explainable with the cash holding behavior before the financial crisis. Confirming previous literature, I find that abnormal cash holding during the financial crisis indeed increased for both U.S. and Continental European firms. Furthermore, I find that during the post-crisis era since 2010, U.S. cash holdings strongly exceed the values predicted based on the pre-crisis time period. For European firms, the increases in mean cash-to-assets ratios after the crisis can be largely explained with the pre-crisis model. I find evidence that the sensitivity of U.S. cash holding to cash flow riskiness increased strongly in the 2010s and can hence be a possible explanation for the observed excess cash values.
6
increased abnormal cash ratios in the crisis period. By using recent data up to 2018, I also provide new insights into more recent dynamics in U.S. and European cash holdings. Specifically, my study provides evidence for a continuation of the U.S. specific trend of increasing cash ratios in the post-crisis period.
The remainder of this paper is structured as follows. I discuss the literature background and formulate the hypotheses in Section 2. In Section 3, I give an overview on the data and related issues. In Section 4, I present the empirical results. Specifically, I discuss a mean and median comparison, display the results from multivariate regressions and a model of excess cash holdings. Finally, I conclude by discussing the findings and managerial implications of this paper in Section 5.
2. Literature background and hypotheses
2.1. European and U.S. cash holding behavior
7
opportunities, many scholars empirically confirm this, see e.g. Opler et al. (1999), Bates et al. (2009) or Fernandes and Gonenc (2016). Furthermore, Bates et al. (2009) show that firm specific attributes changed over time, which led to a larger need for cash holding of U.S. firms in recent decades. They find that in the analyzed time frame of 1980-2006, U.S. firms held less working capital, net of cash, hence less inventories and accounts receivable. Net working capital is seen as a substitute for cash (Opler et al., 1999). Furthermore, Bates et al. (2009) identified the increasing cash flow volatility due to higher idiosyncratic firm risk (Campbell et al., 2001) as a major determinant of rising cash levels, which coincides with precautionary motives for cash holding. As another major determinant, Bates et al. (2009) identify the increased R&D intensity. R&D intensity has been found to be positively related to cash holdings of firms. Reasons for this are the high degree of information asymmetry in R&D investments, as well as the intangible nature of R&D projects (Baldi and Bodmer, 2018). Brown et al. (2009) show that R&D investments are usually financed using cash flow or equity. Since these are highly volatile sources of funding, precautionary cash has to be held in order to ensure a stable financing of R&D projects (Brown and Petersen, 2011). More recently, Lei et al. (2018) provide empirical evidence on the influence of asset tangibility on cash holdings. They argue that firms have to hold more cash when they have less tangible assets as intangible assets cannot easily be used as a collateral for taking a loan.
According to some studies, high cash holding can also be related to agency problems. Following e.g. Jensen (1986), Stulz (1990) and La Porta et al. (2000), managers might have incentives to increase cash ratios in order to obtain private profits. Furthermore, it has been found that high cash holdings might lead to a less efficient use of funds with an extra dollar in cash having a value of less than a dollar for the firm (Dittmar and Mahrt-Smith, 2007; Faulkender and Wang, 2006).
8
firms hold more cash due to tax avoidance reasons. Repatriation of earnings that have occurred elsewhere in the world is usually taxed. However, recent empirical evidence does not seem to support this reasoning. Pinkowitz et al. (2013) do not find a significant effect of the repatriation under advantageous conditions in 2004 in the U.S. Also, Pinkowitz et al. (2016) cannot support that the higher cash holding of U.S. firms compared to other firms in the rest of the world can be traced back to tax reasons.
Next to firm-level characteristics, also country-level characteristics have been found to influence cash holding behavior. For example, previous studies provided evidence regarding the effect of GDP growth, financial development, shareholder and debtor protection or base rates on cash ratios. Similarly to Pinkowitz et al. (2013; 2016), this study focusses on firm-level characteristics and analyzes whether country differences remain after controlling for these. Pinkowitz et al. (2013) use a model of abnormal cash holding and find that increasing cash ratios in the U.S. in the pre-crisis period were a U.S. specific phenomenon. Using a propensity score matching approach, Pinkowitz et al. (2016) find that higher cash ratios in the U.S. are caused by a special type of firm with very high R&D to assets ratios which is only observable in the U.S. Hence, their findings propose that higher U.S. cash holdings are determined by firm-specific characteristics.
9
H1a. European firms hold as much cash as U.S. firms do after controlling for firm-level
characteristics.
H1b. European firms hold less cash than U.S. firms do after controlling for firm-level
characteristics.
Since I do not only analyze European countries as a whole but also separated in country groups, different results for different country groups might occur.
2.2. Dynamics of cash holding behavior in the 21st century
The cash holding behavior in the beginning of the 21st century has been studied by several
scholars. Evidence for increasing U.S. cash ratios until the financial crisis has for example been provided by Bates et al. (2009) or Pinkowitz et al. (2013). Pinkowitz et al. (2013) calculate abnormal cash holdings as the difference between cash levels predicted with a model that is determined based on data from the late 1990s and actual cash to assets ratios. Regarding the pre-crisis period, they find that U.S. firms had higher abnormal increases of cash holdings than firms in other developed countries, which can be traced back to R&D intensive U.S. multinationals’ cash holding behavior.
10
increased during the financial crisis. This is also supported by Pinkowitz et al. (2013). Regarding the financial crisis, they find an increase in abnormal cash levels, especially for highly profitable firms. Pinkowitz et al. (2013) explain this finding with the lack of investment opportunities for these firms.
Bliss et al. (2015) analyze the effect of the financial crisis on corporate payout. They find that a credit supply shock leads to firms decreasing their payouts to investors and therefore retaining more cash in the company. This additionally retained cash is used as a substitute for external financing of investments, but also to increase precautionary cash holdings. These precautionary cash holdings ensure companies to be able to undertake investment opportunities when they arise in the future.
Bliss et al. (2015), however, also mention that the financial crisis was not only associated with a supply shock but also with a demand shock caused by a general reduction in wealth. A demand shock lowers investment opportunities and thus the need for cash. Since cash holding has been found to be connected with agency costs, as discussed above, cash ratios should be lowered following a demand shock (Bliss et al., 2015). Furthermore, Kahle and Stulz (2013) find that only levered firms that depend on bank financing hoarded more cash during the financial crisis. Based on previous literature, I raise the following set of hypotheses regarding cash holding during the financial crisis.
H2a. During the financial crisis, firms hold more cash than predictable based on the
pre-crisis cash holding behavior.
H2b. During the financial crisis, firms did not hold more cash than predictable based on
the pre-crisis cash holding behavior.
11
indication on the issue of European cash holding. Hence, I formulate the following two hypotheses:
H3a. The cash ratio of European firms is differently affected by the financial crisis
compared to that of U.S. firms.
H3b. The cash ratio of European firms is in the same way affected by the financial crisis
compared to that of U.S. firms.
Here again, it is important to consider that European countries of course are very diverse, which is why I am analyzing groups of these countries separately.
The development of cash holding behavior in the aftermath of the financial crisis has not been subject to many studies yet. Since previous literature mostly suggests a positive impact of the crisis, it could be assumed that the post-crisis cash holding moves back to pre-crisis levels and is hence again predictable with the pre-crisis cash holding behavior. However, as cash ratios have been increasing since the 1980s in the U.S. (Bates et al., 2009), one could also assume that this trend continues in the 2010s. My focus again lies on the comparison of U.S. and European cash holding behavior. Thus, I formulate the following hypotheses:
H4a. U.S. abnormal cash holdings are higher than those of European firms in the
post-crisis period.
H4b. U.S. abnormal cash holdings do not differ from those of European firms in the
post-crisis period.
3. Data
12
countries in two major sub-groups: European Anglo-Saxon countries including the U.K., Ireland as well as the Netherlands versus the Continental European countries including France, Spain, Portugal, Italy, Greece, Germany, Austria, Switzerland, Denmark, Sweden, Norway, Finland, Belgium and Luxemburg. I use this separation as the countries around the U.K. are traditionally more similar to the U.S. in their behavior than other European countries (Pinkowitz et al., 2013). Although located on Continental Europe, the Netherlands are considered as Anglo-Saxon as its listed companies are more similar to those of the U.K. than those of elsewhere. All regressions are also run with the Netherlands as part of the Continental European countries and the main inferences are not affected. Based on this definition, I create the binary variables
Anglo Saxon European and Continental European which are dummies equal to one if the firm
is headquartered in the respective region.
The data is provided by Compustat North America and Compustat Global. In case market value data was not provided by Compustat, Thomson Reuters Datastream was used. I gather firm-level data for all available firms. As in previous literature, I exclude financial and utility firms due to their high degree of regulation and special business model (SIC industry codes starting with 49 and 6) and exclude firms with total assets of less than 5 Million USD, e.g. Pinkowitz et al. (2016). Furthermore, I exclude firm-year observations with missing values for any variable. All variables are winsorized at the 1% level. Total assets reported in other currencies were converted into USD using the average yearly exchange rate, obtained from the OECD database. The full dataset contains 81,084 firm-year observations from 11,781 different firms.
13
Table 1. Definition and data source of firm-level variables. Letters in brackets are Compustat IDs. If not specified differently, the data source is Compustat Global for European firms and Compustat North America for U.S. firms.
Variable Definition
Cash Cash and cash equivalents (CHE) divided by the book value of assets (AT).
Size The natural logarithm of total book value of assets (AT), converted to USD based on average year exchange rates, obtained from the OECD database.
Dividends Binary variable equal to 1 if firm pays dividends in the respective year (DVC).
CashFlow Operating income before depreciation (OIBDP), minus interest and related expenses (XINT), income taxes (TXT) and dividends (DVC), standardized by the book value of assets (AT).
CashFlowVol The standard deviation of CashFlow is calculated for each firm and each year for the previous ten years (starting in 1992 for 2002). At least five observations are required to compute a standard deviation. Then, the year averages across two-digit SIC industry codes are computed.
Acquisitions Ratio of acquisitions (AQC) to the book value of assets (AT), missing values are set to zero.
R&D Ratio of R&D expenses (XRD) to the book value of assets (AT), missing values are set to zero.
CaptX Ratio of capital expenditures (CAPX) to the book value of assets (AT).
Leverage Long-term debt (DLTT) plus debt in current liabilities (DLC), standardized by the book value of assets (AT).
NWC Current assets (ACT) net of cash (CHE), minus
current liabilities (LCT), standardized by the book value of assets (AT).
MB ratio Market value of common shares, plus total assets (AT) minus book value of common equity (CEQ), divided by the total book value of assets (AT).
U.S. firms:
The market value is calculated as fiscal year end price of equity (PRCC_F) multiplied by shares outstanding (CSHO).
Western European firms:
14
Table 2. This table presents the summary statistics on firm-level. Sample period is 2002-2018. The data refers to firms headquartered in the U.S. or in Europe (U.K., Ireland, Portugal, Spain, France, Italy, Greece, Belgium, the Netherlands, Luxemburg, Germany, Austria, Switzerland, Denmark, Sweden, Norway and Finland). It is derived from Compustat North America and Compustat Global. In case market value data was not provided by Compustat, Thomson Reuters Datastream was used. All variables are winsorized at the 1% level. Total Assets is given in Million USD. Firms with total assets of less than 5 million USD, as well as utility and financial firms, are excluded. All variables are defined as in Table 1.
Variable N Mean St.Dev Median
Cash 81,084 0.193 0.223 0.106 Total Assets 81,084 3,165.964 9,240.085 345.324 Size 81,084 5.910 2.162 5.844 R&D 81,084 0.052 0.118 0.000 CaptX 81,084 0.048 0.057 0.030 Leverage 81,084 0.230 0.231 0.187 MB ratio 81,084 1.968 1.510 1.476 NWC 81,084 0.027 0.202 0.026 Acquisitions 81,084 0.021 0.057 0.000 Div 81,084 0.474 0.499 0.000 Cashflow 81,084 -0.009 0.256 0.059 CashflowVol 81,084 0.082 0.044 0.071
Fed Rate (in %) 81,084 1.506 1.615 1.127
GDP Growth (in %) 81,084 1.935 1.766 2.250
15
Pinkowitz et al. (2016) report for advanced countries. There are two reasons for this. First, their dataset ends in 2011 and cash ratio means can be assumed to have increased in recent years as Figure 1 depicts. Second, U.S. firms account for more than two thirds of the observations in my dataset. As shown in Table 3 and also found by Pinkowitz et al. (2016), the U.S. mean cash ratio is higher than it is in other countries. The mean firm in my sample has total assets of 3.2 Billion USD and the median firm of only 345.3 Million USD. In my regressions, I follow previous literature and use the logarithmic total assets which have a nearly symmetric distribution. It is furthermore interesting to observe that the median for R&D, Acquisitions and
Div is zero, meaning that there are large numbers of firms that do not spend anything for R&D
and acquisitions or do not pay dividends. The average Fed rate in my sample is 1.506% and the countries grew on average by 1.935 % per year.
16
Table 3: The table presents the summary statistics with number of firm-year observations (N) per country. For R&D, Acquisitions and Div means are reported, for all other variables median values are depicted. Variables are defined as in Table 1. The variable European refers to all considered European countries, listed in single entries below. Anglo (incomplete
term) refers to Anglo-Saxon European countries (GBR, IRL and NLD). Continental European refers to all European countries besides GBR, IRL and NLD listed below.
ISO country code
17
Table 4. This table presents the pairwise correlations between all variables with stars indicating significance at the 1% level. The correlations are computed based on a dataset of 81,084 firm-year observations for the years 2002-2018. All firm-level variables are defined as in Table 1. The data was derived from Compustat North America and Global. In case market value data was not provided by Compustat, Thomson Reuters DataStream was used.
18
The precautionary motive for cash holding could offer a first explanation for higher median and mean cash holdings of U.S. firms, as also discussed by Pinkowitz et al. (2016). The difference in the ratio of dividend paying firms is also striking. While in the U.S. only 28% of the firms are dividend payers, in Europe 87% of the firms are paying dividends. Dividend payments are cash reducers (Pinkowitz et al., 2016) and could thus also explain the higher average U.S. cash holding. Regarding GDP growth, the highest median can be observed for Ireland, with a median rate above 5%. Greece and Portugal display the lowest GDP growth in the considered time frame with median values below 0.9%.
Table 4 depicts the pairwise correlations for all variables. Most correlations are significant at the 1% level but no problematic correlations between independent variables are prevalent. To ensure that collinearity is not biasing the results, I check variance inflation factors (VIF) for each conducted regression. Of course, the binary variables Anglo-Saxon European and
Continental European are highly correlated with the European dummy and will thus not be
19
4. Results
In this section, I investigate whether the cash holding behavior of European firms differs from that of U.S. firms in the considered sample period and whether these differences can be explained with firm-level variables, hence, differences in firm characteristics. First, results of a mean and median comparison are presented. Second, with the means of multivariate fixed effects regressions, I analyze whether firm characteristics are able to explain the differences. Then, I deploy a model of abnormal cash holdings to further investigate how cash holding developed over time during the financial crisis and the post-crisis era.
4.1. Mean and median comparison
In the following, I analyze the cash ratio of U.S. firms in comparison to that of European firms without controlling for firm-level characteristics. Following Pinkowitz et al. (2016), I conduct a mean and median comparison. Table 5 depicts the means and medians for U.S. observations, Continental European observations and Anglo-Saxon European observations. I test for differences in the mean with a two-tailed t-test and for differences in the median with a Wilcoxon rank-sum test. I split up the sample in three time periods: pre-crisis, crisis and post-crisis. With the definition of the crisis I follow previous studies (Kahle and Stulz, 2013; Jang, 2017). The results for an alternative crisis specification of 2009-2010 are depicted in Table A1 in the appendix. This alternative definition results in higher mean and median values during the crisis period reflecting the fact that average cash ratios in 2010 were higher than in 2008 in all considered regions.
20
Table 5. This table presents the means and medians of the cash-to-assets ratio for U.S. firms in comparison to two different groups of European countries, split up in pre-crisis, crisis and post-crisis periods. 14 countries are considered in the group Continental European (France, Germany, Spain, Italy, Greece, Portugal, Austria, Switzerland, Belgium, Luxemburg, Denmark, Sweden, Norway and Finland). Anglo-Saxon European countries includes three countries (U.K., Ireland and the Netherlands). The ratio is calculated as U.S. mean/median divided by the mean/median of the respective region. For means, a two-tailed t-test is conducted to determine whether the means significantly differ. The standard-errors are calculated per time-period and country group. Wilcoxon rank-sum tests were conducted to test for differences in medians per country group and time period. Significance is indicated as *** for the 1% level. All variables are winsorized at the 1% level, firms with total assets of less than 5 Million USD as well as utility and financial firms are excluded.
U.S. Continental European (N = 14)
Anglo-Saxon European (N = 3)
Mean Mean Difference Ratio Mean Difference Ratio
PreCrisis: 2002-2007 0.218 0.135 0.083*** 1.613 0.134 0.084*** 1.629 Crisis: 2008-2009 0.212 0.131 0.081*** 1.617 0.117 0.095*** 1.813 PostCrisis: 2010-2018 0.226 0.140 0.087*** 1.621 0.125 0.101*** 1.808 Median Median Difference Ratio Median Difference Ratio
PreCrisis: 2002-2007 0.117 0.086 0.031*** 1.360 0.079 0.038*** 1.481 Crisis: 2008-2009 0.119 0.090 0.029*** 1.322 0.075 0.044*** 1.587 PostCrisis: 2010-2018 0.126 0.103 0.023*** 1.223 0.078 0.048*** 1.615
21
lower than the pre-crisis level. The findings regarding the comparison of cash ratios could of course be due to differences in firm characteristics and hence do not have to indicate differences in the cash holding behavior. Hence, I will use different approaches to further analyze this issue in the next sections.
4.2. Multivariate regression analysis
4.2.1. Baseline model
22
where all variables are as defined in Table 1. The subscript 𝑗 denotes the industry and 𝑐 the country. Since cash levels and other firm-level variables such as leverage and investment are often seen as jointly determined, see e.g. Opler et al. (1999), I use the 1-year lag of explanatory variables to prevent inconsistent estimations. Hence, for the following regressions cash data for 2003-2018 and data on firm characteristics for 2002-2017 is used. As a robustness check, all regressions are also estimated for the full datasets without lag but the findings do not severely differ.
4.2.2. Expected relation between explanatory variables and cash
This section discusses the expected influence of the explanatory variables included in the baseline regression model on the cash-to-assets ratio.
Continental European (-/n.s.). The dummy variable for firms headquartered in the considered Continental European countries is expected to be either significantly negative or not significant. An insignificant result would indicate that the firm-level characteristics included in the model are able to fully explain the lower average cash holding of Continental European firms. A negative coefficient would indicate that Continental European firms hold less cash than U.S. firms, even after controlling for firm characteristics.
Anglo-Saxon European (-/n.s.). For this indicator variable the same results as for Continental European firms are expected as mean and median ratio are also significantly lower than for U.S. cash holdings. However, the effect might be even stronger since mean and median cash ratios are lower compared to Continental European firms.
23
R&D (+). Larger R&D-to-assets ratios have been found to lead to higher cash-to-assets ratios. This is because R&D investments are seen as intangible and leading to higher levels of information asymmetry (Opler et al., 1999). Furthermore, they are viewed as a proxy for growth opportunities and lead to higher costs of financial distress (Bates et al., 2009).
CaptX (-). If additional assets are obtained that might be used as a collateral, debt capacity increases and, hence, capital expenditures may decrease cash ratios (Bates et al., 2009). Furthermore, Riddick and Whited (2009) find that productivity shocks that increase investment levels may induce higher spending for investment and lower savings of cash, which would also reduce cash levels.
NWC (-). Working capital, net of cash, is seen as a substitute for holding cash and therefore a negative relationship is expected (Bates et al., 2009).
Leverage (+/-). As to leverage, both theoretical explanations for a positive and a negative relationship have been found. Bates et al. (2009) argue that firms will use cash to lower debt if the debt is constraining. According to Ferreira and Vilela (2004), firms with higher leverage ratios are able to raise capital easier and thus need less cash. This would indicate a negative relationship. In contrast, Acharya et al. (2007) et al. argue that higher leverage may be associated with higher cash holding as firms may increase their future funding capacity through the issuance of debt to increase cash or by using cash to reduce debt.
Acquisitions (-). As acquisition activities are often related to cash outflows and are similar in nature to capital expenditures a negative relationship with cash can be assumed (Bates et al., 2009).
24
CashflowVol (+). The industry volatility of cash flows is expected to be positively associated to the cash-to-assets ratio. As discussed by Iskandar-Datta and Jia (2012), a higher volatility of cash flows increases the risks of cash shortfalls. This might induce costs of being financially constrained and thus firms need to hold more cash for precautionary reasons.
Dividends (-). Bates et al. (2009) argue that dividend paying firms are assumed to be able to access capital markets more easily and hence be less financially constrained. Furthermore, these firms are assumed to be less risky. Therefore, the dividend dummy is expected to have a negative sign.
MB ratio (+). The market-to-book ratio is considered as a proxy for growth opportunities, hence higher information asymmetries and higher costs of external funds (Iskandar-Datta and Jia, 2012). Furthermore, high market-to-book ratio firms have more valuable investment opportunities and hence value cash higher (Bates et al., 2009). Thus, high market-to-book ratio firms should have higher cash holdings.
GDP Growth (+/-). GDP growth might be positively associated with cash holdings as companies do not want to miss valuable investment opportunities during times of growth (Fernandes and Gonenc, 2016). However, during times of economic slowdown, as e.g. in the financial crisis of 2008/2009, uncertainty increases and costs for obtaining external financing rise (Gorton, 2009). As argued by Harford et al. (2014), firms will increase cash holding in order to be able to finance projects when external financing is more costly. This would indicate a negative relationship in the sense that negative GDP growth might be associated with higher cash holdings as well.
Fed Rate (+/-). Classical models predict that cash levels fall with increasing interest
25
funds rate, on cash between 1970 and 2014 and find evidence for a robust positive relationship. Therefore, the link between the Federal Reserve fund rate and cash is an empirical issue.
4.2.3. Multivariate results
Table 6 depicts the results for the multivariate regressions based on equation 1, defined above. The bivariate comparison including fixed effects in column 1 confirms previous literature as well as the findings from the mean and median comparison. Firms headquartered in the considered European countries hold on average 6.31% less cash in relation to assets than U.S. firms do. More interestingly, the outcomes presented in column 2 suggest that even after controlling for firm-level characteristics, as well as GDP growth and Federal Reserve funds rate, European firms still seem to have significantly lower cash holdings than U.S. firms. This could indicate that other country-level factors such as e.g. institutional or cultural differences affect the cash holding of Western European and U.S. firms. However, the coefficient in column 2 is much weaker compared to the direct findings in column 1 (0.0198 in comparison to -0.0631). Thus, a large part of the lower European cash holdings can indeed be explained by the considered firm and country-level characteristics. Nevertheless, a difference of -0.0198 in the cash-to-assets ratio may still be considered economically significant as it makes up almost 20% of the median cash holding (0.0198/0.106=0.1868). As mentioned above, I consciously use the median value for the comparison as it is a better representation of the typical firm due to the right-skewness of the cash ratio distribution.
26
Table 6. This table presents results from multivariate linear regressions of firm characteristics and regional dummies on the cash-to-assets ratio based on equation 1. Firms from the U.S. as well as 17 European countries are included. European countries are grouped in Anglo-Saxon (GBR, NLD, IRL) and Continental European (FRA, BEL, LUX, DEU, AUT, CHE, DNK, SWE, NOR, FIN, ESP, ITA, GRC, PRT). Explanatory variables are lagged by one period to minimize an endogeneity bias. Robust standard errors in parentheses are clustered on firm-level. Significance levels are denoted by *** for p<0.01, ** for p<0.05 and * for p<0.1. Variable definitions are given in Table 1. The data was derived from Compustat Global and Compustat North America. In case market value data was not provided by Compustat, Thomson Reuters DataStream was used. All variables are winsorized at the 1% level, firms with total assets of less than 5 Million USD as well as utility and financial firms are excluded.
Dependent variable: Cash (1) (2) (3) (4)
VARIABLES European -0.0631*** -0.0198*** (0.00374) (0.00296) Anglo-Saxon European -0.0728*** -0.0362*** (0.00509) (0.00366) Continental European -0.0572*** -0.00957*** (0.00424) (0.00323) Size -0.00555*** -0.00558*** (0.000689) (0.000688) R&D 0.636*** 0.635*** (0.0202) (0.0202) CaptX -0.453*** -0.455*** (0.0177) (0.0176) Leverage -0.240*** -0.240*** (0.00704) (0.00705) NWC -0.154*** -0.157*** (0.00771) (0.00774) Acquisitions -0.305*** -0.302*** (0.0107) (0.0107) MB Ratio 0.0199*** 0.0199*** (0.00105) (0.00105) Div -0.0353*** -0.0353*** (0.00295) (0.00295) Cashflow -0.0339*** -0.0329*** (0.00895) (0.00896) CashflowVol 0.611*** 0.616*** (0.0360) (0.0360) GDP Growth -0.00123*** -0.000862** (0.000358) (0.000356) Fed Rate -0.00284*** -0.00300*** (0.000495) (0.000494) Constant 0.165*** 0.218*** 0.166*** 0.217*** (0.0264) (0.00674) (0.0268) (0.00674) Observations 66,312 66,312 66,312 66,312 R-squared 0.237 0.496 0.237 0.497
Year Fixed Effects Yes No Yes No
27
While the coefficient for Anglo-Saxon European countries is larger than that for Europe as a whole, the coefficient for the Continental European countries is much smaller. However, the value of -0.00957 is nearly 10% of the median cash ratio and may thus still be considered economically significant (0.00957/0.106=0.0903).
Regarding the firm-level characteristics, the findings confirm the expectations formulated in Section 4.2.2. above. The market-to-book ratio as well as the R&D-to-assets ratio as proxies for firms with growth opportunities and higher degrees of information asymmetry positively influence the cash ratio. Also, industry cash flow volatility has a significantly positive influence on cash holding, confirming that firms need to hold more cash for precautionary reasons if the cash flow risk is higher. By contrast, firm size, capital expenditures, net working capital, acquisitions and dividends negatively impact the cash ratio. Regarding leverage, I find a negative relationship, supportive of the argumentation by Bates et al. (2009) that firms use cash if debt is constraining. A higher cash flow is associated with lower cash holding which indicates that cash flow may indeed be used as a substitute for holding cash.
Furthermore, I find that GDP growth is negatively associated with the cash-to-assets ratio. This indicates that after controlling for growth opportunities with other variables firms indeed seem to hold more cash in times of weaker economic growth as e.g. during crises, which has also been found by previous literature. This could be due to higher cost for external financing (Harford et al., 2014). Regarding the Fed Rate, I find a negative relationship with cash, which supports classic theories of cash holding that predict that higher interest rates induce higher opportunity costs for cash holding. Hence, the low values of interest rates since the financial crisis can be assumed to be associated with increased cash holding of firms.
28
qualitatively the same. Additionally, I run the regressions without lag to analyze the robustness of my findings. The results depicted in Table A2 in the appendix are very similar to those including the lag. Interestingly however, the GDP growth is not significant in the non-lagged model. Hence, the negative effect of GDP growth on cash seems to only occur one year later and not in the same year. An explanation for this might be that firms need some time to adjust their cash ratio to differing refinancing costs and also to build up slack in their cash holding. As a further robustness check, I run the regressions with the Netherlands included in the group of Continental European instead of Anglo-Saxon European firms, shown in Table A3 in the appendix. I find that all results are not changing qualitatively. The coefficient for Anglo-Saxon European firms is slightly more negative and the coefficient for Continental European firms is slightly more positive. Hence, the Dutch cash holding seems to be similar to both regions in some way.
29
relationship between firm characteristics and cash holding, I conduct separate regressions for all of the three regions in the next section.
4.2.4. Separate analysis of country groups
To analyze how the influence of firm-level characteristics on cash holding behavior may vary between the considered groups of countries, I follow the approach of Iskandar-Datta and Jia (2012) and conduct separate regressions for U.S. firms, Anglo-Saxon European and Continental European firms. The results are presented in Table 7. It can be observed that the sign of the coefficients is the same for all firm-level characteristics. This indicates that in the considered time frame the directions of the firm attributes’ effects on Cash were the same for U.S., Continental European and Anglo-Saxon European firms alike. The magnitude of the effect, however, varies strongly for some of these attributes, which is in line with the findings of Iskandar-Datta and Jia (2012). Specifically, the magnitude of the effects of R&D expenses, capital expenditures, acquisitions, dividends, cash flow and the industry cash flow volatility on the cash-to-assets ratio seems to differ. Mostly, the coefficients for Anglo-Saxon European countries are more similar to those of the U.S. than the coefficients in the Continental European model, which might be a result of the higher proximity of especially the U.K. with the U.S, as also discussed by Pinkowitz et al. (2013). The most striking differences can be observed for
CashflowVol, where the coefficient for U.S. firms is more than five times higher than for
30
Table 7. This table presents results from multivariate linear regressions of firm characteristics on the cash-to-assets ratio. Regressions are run separately for firms headquartered in the U.S in Anglo-Saxon Europe (GBR, NLD, IRL) and in Western Continental Europe (FRA, BEL, LUX, DEU, AUT, CHE, DNK, SWE, NOR, FIN, ESP, ITA, GRC, PRT). Explanatory variables are lagged by one period to minimize an endogeneity bias. Robust standard errors in parentheses are clustered on firm-level. Significance levels are denoted by *** for p<0.01, ** for p<0.05 and * for p<0.1. Variable definitions are given in Table 1. The data was derived from Compustat Global and Compustat North America. In case market value data was not provided by Compustat, Thomson Reuters DataStream was used. All variables are winsorized at the 1% level, firms with total assets of less than 5 Million USD as well as utility and financial firms are excluded.
Dependent variable: Cash (1) (2) (3)
VARIABLES U.S. Continental
European Anglo-Saxon European Size -0.00452*** -0.00891*** -0.00563*** (0.000968) (0.00101) (0.00129) R&D 0.628*** 0.466*** 0.551*** (0.0220) (0.0684) (0.0792) CaptX -0.526*** -0.182*** -0.230*** (0.0205) (0.0401) (0.0420) Leverage -0.242*** -0.254*** -0.211*** (0.00813) (0.0142) (0.0175) NWC -0.157*** -0.146*** -0.147*** (0.00966) (0.0143) (0.0209) Acquisitions -0.326*** -0.179*** -0.216*** (0.0126) (0.0228) (0.0265) MB Ratio 0.0204*** 0.0176*** 0.0194*** (0.00119) (0.00276) (0.00268) Div -0.0355*** -0.0163** -0.0647*** (0.00353) (0.00640) (0.00900) Cashflow -0.0252** -0.0722** -0.0882*** (0.00980) (0.0287) (0.0315) CashflowVol 0.812*** 0.161*** 0.344*** (0.0489) (0.0541) (0.0671) GDP Growth -0.00181*** -0.00250*** 0.00187** (0.000541) (0.000557) (0.000727) Fed Rate -0.00429*** -0.000543 -0.00108 (0.000637) (0.000879) (0.00106) Constant 0.202*** 0.239*** 0.207*** (0.00867) (0.0112) (0.0136) Observations 46,149 12,509 7,654 R-squared 0.508 0.309 0.359
31
Europe, another explanation raised by Iskandar-Datta and Jia (2012) based on the differing financing systems might be suitable. The U.S. is considered to have a market-based system where the major source of external funding are financial markets.
In contrast, Continental Europe, especially Germany, France and Italy, is classified to have a bank-based system with banks being the most important capital providers. As stated by Iskandar-Datta and Jia (2012) a bank-based system leads to reduced information asymmetry in external funding. Therefore in situations of high uncertainty where information asymmetries are increased, as measured by higher cash flow volatility or higher R&D expenses, Continental European firms might need to hold less cash in comparison to U.S. firms because they are able to obtain financing through their house bank.
32
Interestingly, I find that U.S. firms decrease their cash ratio more per additional unit of capital expenditures than European firms do. A reason for this could be that firms having high cash balances in Europe were also able to invest more during the sample period, which weakens the negative relationship. Especially during the financial crisis, ensuring the ability to invest was an important motive for holding cash (Bliss et al., 2015). Regarding the effect of being a dividend payer, the two groups of European countries differ strongly. In the Anglo-Saxon European countries being a dividend paying firms decreases cash holding more than in the U.S. In Continental Europe, being a dividend paying firm decreases cash ratios less than in the U.S. Iskandar-Datta and Jia (2012) even find a positive link between the dividend dummy and cash ratios for France and Germany. They explain this with the bank-based system that allows non-dividend paying financially-constrained firms to hold less precautionary cash. This could also drive the lower decrease in cash ratios for dividend payers in Continental Europe. An explanation could be that absolute dividend payments are larger in the U.S. and Anglo-Saxon European countries compared to Continental Europe. However, a two-tailed t-test on mean equality of the acquisition to assets ratio does not indicate significant differences between the two groups with only dividend paying firms being considered.
33
for all three regions but only significant for the U.S. Since the Federal Reserve is a U.S. institution this is not surprising.
Overall, the cash holding behavior of U.S., Continental European and Anglo-Saxon European firms seems to be determined by the same firm characteristics. While the direction of the effect is the same across all considered country groups, the magnitudes of the effects vary. Especially, the industry cash flow volatility has a much stronger negative impact on cash holding in the U.S. than in both Western European regions. Three non-mutually exclusive explanations may help to clarify this issue. The cultural explanation might be able to explain lower Anglo-Saxon cash holding, as especially the UK and Ireland score lower than the U.S. on uncertainty avoidance in the classification by Hofstede (2001). However, for Continental Europe this does not hold. There, the bank-based system might induce lower information asymmetries and therefore a lower need for cash in relation to cash flow uncertainty. The strategical focus of U.S. firms on ensuring stable financing of R&D investments could explain both lower Continental European and lower Anglo-Saxon European cash holdings in relation to higher cash flow uncertainty.
34
effects and excluding the GDP Growth and Fed Rate. The results again do not change qualitatively.
4.2.5. Interaction term analysis
The separated regression analyses of U.S., Continental European and Anglo-Saxon European firms in the previous section have shown that the magnitude of the effect of certain firm characteristics on the cash-to-assets ratio varies strongly. A simple multivariate regression model, such as the one used above, does not allow for such variations, as discussed by Pinkowitz et al. (2016). Therefore, I deploy a model including interaction terms of the variables that showed the most strongly differing magnitudes with indicator variables for Anglo-Saxon European and Continental European firms and investigate whether lower cash holdings of the European firms can be explained.
The regression results are presented in Table 8. Column 1 displays the model including the interaction of CashflowVol with the regional dummies. As discussed above, the influence of
CashflowVol on the cash ratio showed the strongest variability across Western European and
U.S. firms. I find positive and significant coefficients for the interaction of the Anglo-Saxon European as well as the Continental European indicator variable with CashflowVol. These coefficients may also be considered as economically significant. A one standard deviation increase of CashflowVol for Continental European (Anglo-Saxon European) firms decreases the cash-to-assets ratio by 0.0311 (0.0175) less than for the U.S. firms and makes up 29.33% (16.51%) of the median cash ratio. This is not surprising, as Table 7 already showed that U.S. firms increase cash holding more for each unit of cash flow risk than Western European firms. Therefore, the focus of this analysis lies on the regional dummies.
35
Table 8. This table presents results from multivariate linear regressions of firm characteristics, regional dummies and interaction terms on the cash-to-assets ratio. Firms from the U.S. as well as 17 European countries are included. European countries are grouped in Anglo-Saxon European (GBR, NLD, IRL), abbreviated by Anglo-Saxon, and Continental European (FRA, BEL, LUX, DEU, AUT, CHE, DNK, SWE, NOR, FIN, ESP, ITA, GRC, PRT). Explanatory variables are lagged by one period to minimize an endogeneity bias. Robust standard errors, in parentheses, are clustered on firm-level. Significance levels are denoted by *** for p<0.01, ** for p<0.05 and * for p<0.1. Variable definitions and data sources are given in Table 1. All variables are winsorized at the 1% level, firms with total assets of less than 5 Million USD as well as utility and financial firms are excluded.
Dependent Variable: Cash (1) (2) (3) (4)
VARIABLES Anglo-Saxon -0.00534 -0.0342*** -0.0513*** -0.0205*** (0.00638) (0.00369) (0.00458) (0.00656) Continental European 0.0414*** -0.00288 -0.0271*** 0.0237*** (0.00542) (0.00333) (0.00424) (0.00594) CashflowVol x Anglo-Saxon -0.398*** -0.401*** (0.0824) (0.0827)
CashflowVol x Continental European -0.708*** -0.616***
(0.0688) (0.0699)
R&D x Anglo-Saxon -0.0577 0.0397
(0.0795) (0.0851)
R&D x Continental European -0.312*** -0.173**
(0.0618) (0.0694)
CaptX x Anglo-Saxon 0.317*** 0.312***
(0.0472) (0.0464)
36
Although the coefficient for Anglo-Saxon European countries is still negative, it is no longer significant. Hence, the significantly lower cash holding of Anglo-Saxon European firms can be fully explained by the differing effect of CashflowVol. Regarding Continental European firms, I even find a significantly positive coefficient on the regional dummy, indicating that these firms hold more cash than U.S. firms when controlling for both, the differing effect of cash flow uncertainty and the firm characteristics.
Column 2 of Table 8 presents the results for the incremental effect of R&D on Anglo-Saxon European and Continental European firms. For both country groups, I find negative coefficients on the interaction terms, but the effect is only significant for Continental European firms. Hence, Anglo-Saxon European firms do not hold significantly more or less cash per unit of
R&D as compared to U.S. firms. Controlling for these effects, I find that the dummy for firms
headquartered in Continental Europe is insignificant. Hence, while not being as relevant for the Anglo-Saxon European countries, the differing incremental effect of R&D expenses is also able to explain the lower cash holding of Continental European firms in comparison to U.S. firms. In column 3, I control for the differing effect of capital expenditures on cash. As discussed in the previous section, the Western European firms hold more cash per unit of capital expenditures than U.S. firms – in contrast to R&D and CashflowVol. In line with the previous findings, the coefficients are positive and significant for both groups of Western European countries. Consequently, the dummy variables for both country groups are more negative compared to the baseline model. Hence, accounting for the differing magnitude of the effect of capital expenditures on cash even increases the non-explainable difference in cash holdings between both groups of Western European firms and U.S. firms.
37
interaction with R&D in column 2. The same holds for the incremental effect of CashflowVol on the cash ratio. Thus, the incremental effects of R&D expenses and industry cash flow volatility for Continental Europe seem to overlap. This supports the differing strategy explanation raised in the previous section, according to which U.S. firms hold more cash per unit of cash flow volatility in order to ensure their R&D leadership. The coefficient on the interaction variable of R&D and the Anglo-Saxon European dummy is positive, however, still insignificant. Regarding capital expenditures, the incremental effect for Continental Europe is slightly weaker as compared to column 3. After controlling for the differing relationship of
CashflowVol, R&D and CaptX on Cash, I again find that Continental European cash ratios are
significantly larger than those of U.S. firms. Thus, the previously indicated lower cash holding of Continental European firms can be fully explained with the differing effect of cash flow volatility and R&D expenses. Instead, even a positive unexplainable difference remains. Iskandar-Datta and Jia (2012) also find that, after controlling for firm characteristics, German firms which make up most of the Continental European observations, hold too much cash which they attribute to possible agency problems.
Regarding Anglo-Saxon European firms, I still find a negatively significant coefficient which is, however, weaker than in the baseline model. Thus, the differing effects of firm attributes are also able to partly explain the lower cash holding of the Anglo-Saxon European firms. The remaining unexplainable part in my model might be caused by differing country-level variables such as cultural differences or corporate governance. It might also be explainable by differing effects of other firm-level variables. However, using an interaction between Div and the regional dummies results in collinearity problems due to the low in-sample variation as more than 85% of Western European firms are dividend payers. Also, the incremental effects of
Cashflow and Acquisitions have been analyzed. In unreported results, I find that the interaction
38
previous finding that U.S. firms hold less cash for an additional unit of acquired assets. Including the interaction in the full model does not affect the inferences made above.
The regression results in this section are robust to several checks. First, in unreported results, all regressions are conducted with year fixed effects and excluding GDP Growth and the Fed
Rate. Second, the findings do not qualitatively change when non-lagged explanatory variables
are used, as depicted in Table A7 in the appendix. Third, the regressions are ran with the Netherlands included in the group of Continental European countries and the Anglo-Saxon group only consisting of the U.K. and Ireland, see Table A8 in the appendix. I find that all inferences are robust to this alternative specification. The incremental effect of R&D for Continental European firms is less significant when the Netherlands are included, which indicates that the classification as Anglo-Saxon is indeed more suitable.
39
4.3. Model of excess cash holdings
After having analyzed how and why U.S. and Western European cash holding behavior differs, this section discusses how these differences in cash holding developed over time between 2002 and 2018. Especially, the effect of the financial crisis and the post-crisis period is in scope. In this section, I deploy a model of abnormal cash holdings based on the approach by Pinkowitz et al. (2013). With such an approach, a base period is used to configure a firm cash holding model. This model can then be used for an out-of-sample prediction of cash holding values. By comparing these model-based predictions with the true cash holdings, it is possible to calculate abnormal cash holdings. Hence, excess cash is defined as the difference between predicted and true cash holdings. As stated by Pinkowitz et al. (2013), such a model allows to infer how much more or less cash firms hold than they would have if their cash holding behavior had not changed.
4.3.1. Model definition
In line with Pinkowitz et al. (2013), I use the cash holdings model proposed by Bates et al. (2009) to proxy the different motives of firm cash holding. Specifically, the regression model is defined as 𝐶𝑎𝑠ℎ𝑖𝑡 = 𝛽0+ 𝛽1× 𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽2× 𝑅&𝐷𝑖𝑡+ 𝛽3× 𝐶𝑎𝑝𝑡𝑋𝑖𝑡 + 𝛽4× 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡 + 𝛽5× 𝑁𝑊𝐶𝑖𝑡+ 𝛽6× 𝐴𝑐𝑞𝑢𝑖𝑠𝑖𝑡𝑖𝑜𝑛𝑠𝑖𝑡 + 𝛽7× 𝐶𝑎𝑠ℎ𝐹𝑙𝑜𝑤𝑖𝑡 + 𝛽8× 𝐶𝑎𝑠ℎ𝐹𝑙𝑜𝑤𝑉𝑜𝑙𝑗𝑡 + 𝛽9× 𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑𝑠𝑖𝑡 + 𝛽10× 𝑀𝐵 𝑟𝑎𝑡𝑖𝑜𝑖𝑡 + 𝛾𝑐+ 𝜀𝑖𝑡, (2)
40
holding. I define the base period as 2002-2006. This is the pre-crisis time frame as mentioned above but excluding the year 2007. The financial crisis started in the U.S. in 2007 already and I seek to avoid the crisis effect to be included in the cash holding predictions.
Following Pinkowitz et al. (2013), I conduct a pooled cross-sectional regression on all sample firms for the years 2002-2006 to estimate the model. The estimated coefficients are shown in Table A9 in the appendix. All variables have the expected sign and confirm previous literature, e.g. Opler et al. (1999) or Pinkowitz et al. (2013).
4.3.2. Results
Based on the estimated model from equation 2, I compute out-of-sample predictions for the cash-to-assets ratio of all firms for the years 2007-2018. In the next step, I compute the excess cash, defined as the difference between the predicted and the true cash ratio. A firm that holds more cash than predicted in the respective year thus holds positive excess cash.
41
Table 9. This table presents excess cash holdings, computed as the difference between the predicted cash-to-assets ratio and the true value. The predicted values stem from a regression estimated of firm characteristics on the cash ratio for all firms for the years 2002-2006 including country fixed effects, defined in equation 2. The estimated regression coefficients are given in Table A9 in the appendix. Firms from the U.S. and 17 European countries are included. European countries are grouped in Anglo-Saxon European (GBR, NLD, IRL) and Continental European (FRA, BEL, LUX, DEU, AUT, CHE, DNK, SWE, NOR, FIN, ESP, ITA, GRC, PRT). Standard errors are calculated by year/sub period and country (group). Significance levels are denoted by *** for p<0.01, ** for p<0.05 and * for p<0.1, t-values are shown in parentheses. The data was derived from Compustat Global and Compustat North America. In case market value data was not provided by Compustat, Thomson Reuters DataStream was used. All variables are winsorized at the 1% level and firms with total assets of less than 5 Million USD are excluded.
Panel A: Excess cash holdings by year
U.S. Continental European (N=14) Anglo-Saxon European (N=3) 2007 0.011*** 0.002 -0.002 (3.732) (0.644) (-0.465) 2008 0.010*** 0.003 0.001 (3.234) (0.886) (0.312) 2009 0.006** 0.013*** -0.001 (1.972) (3.577) (-0.274) 2010 0.015*** 0.010*** -0.001 (4.900) (2.711) (-0.144) 2011 0.016*** 0.003 -0.003 (5.123) (0.907) (-0.653) 2012 0.013*** 0.002 -0.010** (4.253) (0.676) (-2.242) 2013 0.019*** 0.002 -0.003 (5.865) (0.507) (-0.552) 2014 0.027*** 0.003 -0.003 (8.539) (0.686) (-0.531) 2015 0.032*** 0.004 0.011** (9.828) (1.149) (1.950) 2016 0.016*** 0.005 0.003 (4.737) (1.379) (0.531) 2017 0.013*** -0.004 0.000 (3.783) (-1.046) (0.042) 2018 0.021*** 0.003 0.005 (5.985) (0.803) (0.860)
Panel B: Excess cash holdings by sub period
Crisis: 2008-2009 0.008*** 0.008*** 0.000
(3.709) (3.047) (0.049)
Post Crisis: 2010-2018 0.019*** 0.003*** 0.000
(17.724) (2.604) (-0.004)
post-42
crisis period seem to be a U.S. specific phenomenon and cannot be explained with earlier literature.
In a first step, I analyze the crisis years. The aggregated results in Panel B of Table 9 suggest that U.S. and Continental European firms held significantly positive excess cash during the period of the financial crisis. With the crisis defined as 2008-2009, I find the same average level of excess cash holding for both regions, which confirms results by Pinkowitz et al. (2013). Considering the yearly excess cash holdings in Panel A, it is observable that the abnormal cash holding of Continental European firms was affected much later than that in the U.S. This, however, is not surprising, as the crisis started in the U.S. already in 2007 and led to a recession already in 2008 while Europe’s economy still grew in 2008. In Table A10 in the appendix, results for a crisis definition of 2009-2010 are depicted. As to be expected, this crisis definition suggests a stronger crisis effect for Continental Europe with aggregated excess cash of 0.12 compared to 0.10 in the U.S.
43
Summarizing, I can confirm hypothesis H2a for Continental Europe and U.S. firms but the findings regarding the Anglo-Saxon European firms are supportive of hypothesis H2b. The crisis effect seems to be of similar magnitude for U.S. and Continental European firms but not for Anglo-Saxon European firms. Therefore, considering hypothesis H3, the findings are supportive of hypothesis H3b for Continental European firms while for Anglo-Saxon European firms hypothesis H3a is supported.
During the post-crisis period, an increase in U.S. excess cash ratios is observable. The mean difference between predicted value and true cash-to-assets ratio for U.S. firms for 2018 is with 0.021 nearly twice as high as for 2007 with 0.011. The aggregated results in Panel B indicate that U.S. firms held on average excess cash of 0.019 in the post crisis period. This refers to nearly 18% of the median cash holding ratio in my sample (0.019/0.106=0.1792).
44
included in the Anglo-Saxon group, the positive excess cash is even stronger in 2015, which also supports the reasoning above (See Table A11 in the appendix).
For 2012, I even find a negatively significant excess cash for Anglo-Saxon Europe. However, as Table A11 in the appendix shows, this seems to be mainly drawn by Dutch firms as the excess cash is no longer significant when the Netherlands are not considered part of Anglo-Saxon Europe. Especially, the larger Dutch firms were able to reduce their working capital balances better than other European firms in the post-crisis period as stated by the PwC working capital report for 2019 (PwC, 2020). In connection with the slight recession that the Netherlands experienced in 2012, this might explain that cash balances fell below the prediction in that respective year.
45
To sum up, I can document that U.S. cash ratios have increased far beyond the values predictable with the early 2000s cash holding behavior, therefore continuing the previous trend of increasing cash ratios since the 1980s as analyzed by Bates et al. (2009). This trend is not observable in Europe where the cash holding patterns from the early 2000s are still able to explain most of the variations in cash ratios. Hence, my findings clearly support my hypothesis H4a. As a possible explanation, the increasing sensitivity of U.S. cash holdings to cash flow riskiness is identified in this study.
In contrast to the U.S. specific phenomenon of increasing cash ratios in the post-crisis period, the magnitude of excess cash observable during the financial crisis is similar in Continental Europe and the U.S. and is in line with theoretical and empirical findings from previous literature. Regarding the crisis, the Anglo-Saxon European firms seem to be in a special role as no significant excess cash is observable. A possible explanation for this might be that the Anglo-Saxon European firms had the most efficient working capital management during the crisis and thus held lower cash balances. This is indicated by a PwC working capital report stating that especially firms in the U.K. were able to hold the lowest working capital balances in relation to sales during the late crisis period (PwC, 2012). Further, Kahle and Stulz (2013) find that only levered firms that are bank dependent increased their cash holding due to the financial crisis. Hence, a reason could be that firms in the Anglo-Saxon European countries are less bank dependent than firms in the U.S. and Continental Europe.
5. Conclusion and managerial implications
46
47
2010s until 2018 in my regressions while the sample of Pinkowitz et al. (2016) only reaches up until 2011. As the excess cash holding analysis shows, U.S. cash holdings were similar to those in Continental Europe during the crisis, but increased far beyond the predictions in the 2010s. In contrast to the excess cash in the crisis period, I find that the post-crisis excess cash holding is a U.S. specific phenomenon. I document an increasing sensitivity of U.S. cash ratios to cash flow uncertainty from the years 2003-2009 to the period 2010-2018 which might cause the strong increase in U.S. excess cash holding. The strategy direction, discussed above, might be able to explain the increasing sensitivity to cash flow volatility. Specifically, the strategical focus on stable funding of R&D might be more distinct since the post-crisis period. The U.S. are known widely for their highly innovative firms such as Google, Amazon or Facebook which also hold high amounts of cash (Pinkowitz et al., 2013; 2016). However, my findings suggest that not only the highest R&D spenders but also a larger share of U.S. firms differs in their cash holding behavior. Future research may further study the higher sensitivity of U.S. firms to cash flow uncertainty and in which way the strategical focus might differ to European firms. Next to differing firms, also the effect of country-level factors such as the corporate governance standards, investor protection or regulation reasons might be studied in this relation.
48
This special role of the U.K. could be subject to future research studying why the British cash ratios do not significantly differ from predictions based on pre-crisis cash holding during the crisis. A possible explanation for this might be based on Kahle and Stulz (2013) according to whom only bank dependent firms need to increase their cash ratio during the time of the crisis.
This work provides several implications that might be relevant for managerial and policy issues. First, I provide further on the effect of the financial crisis on firm cash holding which supports the theoretical reasoning that cash has advantages in times of financial distress and might support managers’ decisions in determining the cash level to be held in times of an economic crisis. Considering the economic effects of the current Corona virus pandemic, my work suggests that firms should further increase abnormal cash holdings. The question whether those firms with higher cash-to-assets ratios before the virus outbreak will have higher chances to survive and be able to grow quicker afterwards will be interesting to study by future research. Second, for U.S. firms my findings suggest that extraordinary high cash holdings, which have been criticized by the public as being costly and leading to lower investments, might in fact be useful in ensuring the innovativeness of U.S. firms. Considering the issue from a policy perspective with the goal to reduce cash-to-assets ratios, my findings suggest that providing alternative ways of funding for industries with high cash flow volatilities may help.
From a European policy perspective, European governments might be interested in fostering more innovative industries such as the technological and pharmaceutical sector as my research shows that there are much less of these firms listed in European countries as compared to the U.S.
49
50
Appendix
Table A1. This table presents the means and medians of the cash-to-assets ratio for U.S. firms in comparison to two different groups of European countries, split up in the alternatively defined pre-crisis, crisis and post-crisis period. 14 countries are considered in the group Continental European (France, Germany, Spain, Italy, Greece, Portugal, Austria, Switzerland, Belgium, Luxemburg, Denmark, Sweden, Norway and Finland). Anglo-Saxon countries includes three countries (U.K., Ireland and the Netherlands). The ratio is calculated as U.S. mean/median divided by the mean/median of the respective region. For means, a two-tailed t-test is conducted to determine whether the means significantly differ. The standard-errors are calculated per time-period and country group. Wilcoxon rank-sum tests were conducted to test for differences in medians per country group and time period. Significance is indicated as *** for the 1% level. All variables are winsorized at the 1% level, firms with total assets of less than 5 Million USD as well as utility and financial firms are excluded.
U.S. Continental European (N = 14)
Anglo-Saxon European (N = 3)
Mean Mean Difference Ratio Mean Difference Ratio
PreCrisis: 2002-2008 0.217 0.134 0.083*** 1.620 0.131 0.085*** 1.648 Crisis: 2009-2010 0.220 0.143 0.078*** 1.544 0.123 0.097*** 1.791 PostCrisis: 2011-2018 0.227 0.139 0.088*** 1.631 0.126 0.101*** 1.804 Median Median Difference Ratio Median Difference Ratio
