Why Are Japanese Firms Still Increasing Cash Holdings?

Why Are Japanese Firms Still Increasing Cash Holdings? Abstract Japanese firms resumed accumulation of cash to the highest cash holding levels among developed economies after the 2008 financial crisis. We investigated the cash saving behavior of these firms over the period 2001 11, focusing on the firm-bank relationship and R&D expenditure. We found that bank-dependent firms have higher cash holdings than those with access to the bond markets. On the other hand, the differences in cash holdings between R&D-intensive firms and non-r&d-intensive firms are larger for firms that have access to the bond markets compared with the bank-dependent ones. We also observed that bank-dependent firms show larger cash flow sensitivity of cash than firms that have access to the bond markets. Furthermore, the effect of R&D on cash flow sensitivity of cash is significant only for firms that have access to the bond markets. These results suggest that bank-dependent firms cannot flexibly change their cash saving policy responding to precautionary demands for cash stemming from R&D investment. JEL Classification: G32 Keywords: Cash policy; cash flow sensitivity; bank power; financing constraint

1. Introduction There are two important aspects pertaining to corporate cash hoarding in the 21 st century. As indicated by Pinkowitz, Stulz, and Williamson (2013), cash holdings of US and Japanese enterprises are significantly higher than that of firms in other developed economies. 1 Bates, Kahle, and Stulz (2009) observed that the cash-to-assets ratios of US firms more than doubled from 1980 to 2006. Bates et al. investigated the factors that determined increased cash holdings among US firms and found that this can be attributed to increasing precautionary demands for cash, mainly triggered by growing demands for R&D. On the other hand, Japanese firms resumed accumulation of corporate cash holdings after the 2008 financial crisis (Figure 1). This is surprising because higher levels of cash holdings among Japanese firms are associated with bank power (Pinkowitz and Williamson, 2001), and bank power is said to be weakening after the deregulation of the bond markets in the 1990s. There are two possible reasons explaining such remarkable cash savings among Japanese firms. First, since cash holdings play an important role in financing R&D 1 In another sample, Iskandar-Datta and Jia (2012) show that the cash-to-assets ratio of Japanese firms is the highest among seven industrialized countries, although the difference between Japan and other countries in this regard shrunk significantly in the 1990s. Rajan and Zingales (1995) report that Japanese companies held larger amounts of cash and short-term investments during the sample period.

projects (Brown and Petersen, 2011), a growth in R&D demand in Japanese firms to gain competitive edge against their foreign rivals would increase precautionary demands for cash. Cash holdings for precautionary motives are especially important in Japan because equity financing is very limited in the Japanese stock market and lines of credits are not common among Japanese companies. 2 In fact, according to a survey of CFOs of Japanese firms (Sasaki, Sasaki, Xu, and Hanaeda, 2014), holding excess cash is important in preparing for future cash flow shortfalls and future investment opportunities. Although recent studies show that non-operational cash guards against future cash flow shocks and lines of credit give firms the option to exploit future business opportunities (Lins, Servaes, and Tufano, 2010), most Japanese firms hold non-operational cash to meet both these objectives (Sasaki et al., 2014). Alternatively, the 2008 financial crisis may see the return of monopoly power of banks against client firms that rely on banks for financing projects. Notably, even after the deregulation of the corporate bond markets in 1990s, access to the bond markets is limited in case of Japanese listed companies. In fact, corporate bond issuances have been 2 According to Sasaki et al. (2014), who surveyed CFOs of Japanese companies, 29% of respondents said that they can issue equity whenever they need to raise capital. Since the respondents belonged to large firms, these results indicate that even large firms face difficulty in raising capital from the equity markets. Moreover, they are not in favor of lines of credits. Around 46% of respondents said that their companies had never used lines of credits. It is also worth noting that firms with access to capital markets use lines of credits unlike firms without them.

limited to firms with high corporate bond grades, and many Japanese firms still rely on bank loans following the deregulation of the bond issuance markets in the 1990s. Pinkowitz and Williamson (2001) present evidence consistent with the argument that high cash holdings in Japan can be attributed to the power of banks. If main banks hold a monopolistic power position, they can extract rents from their client firms by demanding high cash reserves. Although banks hardly require their client firms to accumulate bank deposits directly, client firms may voluntarily increase bank deposits to maintain good relationships with their main banks. Thus, we expect that firms that have access to the bond markets for raising funds can more flexibly change their cash reserve policy based on precautionary demands, such as financing of R&D projects. We verified this hypothesis by investigating the cash flow sensitivities of cash (CFSCs), focusing on firm-bank relationship and financing of R&D investments, using a new data sampled from Japanese firms over the period of 2001 2011. We observed that bank-dependent firms still have higher cash holdings than firms with access to the bond markets. On the other hand, the differences in cash holdings between R&D-intensive firms and non-r&d-intensive firms are larger for firms that have access to the bond markets as against bank-dependent firms. We also find that bankdependent firms show larger cash flow sensitivity of cash, whereas the effect of R&D on

cash flow sensitivity is significant only for bank-dependent firms. These results indicate that firms that have access to the bond markets determine their cash policy based on precautionary demands for cash, whereas bank-dependent firms unanimously accumulate cash holdings to maintain good relations with banks. Our paper contributes to the literature by showing that firms cash-saving behavior differs according to the extent of bank dependence. Especially, we complement the finding of Pinkowitz and Williamson (2001) by showing that bank-dependent firms cannot flexibly change their cash saving policy responding to precautionary demands for cash stemming from R&D investment. The remainder of the paper is structured as follows: Section 2 briefly reviews the literature and develops hypotheses, section 3 describes the data and methodology used, section 4 presents empirical results, and section 5 concludes. 2. Literature and hypotheses Literature argues that the precautionary motive, advocated by Keynes and further formulated by Almeida et al. (2004), plays an important role in explaining the substantial cash holdings of firms in the developed economies. According to the precautionary motive theory, firms increase their cash holdings to prepare for adverse

shocks when access to the capital markets is expensive. For firms facing financing frictions, internal cash flows are the main source used to build sufficient cash holdings. Almeida et al. (2004) model this demand for cash and predict substantial CFSCs for financially constrained firms, while those for unconstrained firms are not significantly different from zero. The authors present evidence consistent with this argument using various criteria for financing friction. 3 Opler et al. (1999) examine the determinants of corporate cash holdings among publicly traded US firms during the 1971 94 period and find that firms with strong growth opportunities and riskier cash flows hold relatively high ratios of cash to total non-cash assets. Furthermore, they observe that firms with the greatest access to the capital markets tend to hold lower ratios of cash to total non-cash assets. Bates et al. (2009) find that the cash-to-assets ratios of US firms more than doubled from 1980 to 2006. Bates et al. investigate the determinants of increased cash holdings among US firms and find that this can be attributed to increasing precautionary demands for cash, mainly triggered by growing demand for R&D. An increasing demand for R&D 3 Almeida et al. (2004) argue that the advantage of examining CFSC instead of cash flow sensitivity of investment (CFI) in scrutinizing the effect of financing friction is circumventing the possibility of private information regarding investment opportunities inducing a superficial relationship between cash flows and investment. This is because while financially constrained firms choose their optimal cash policies to balance the profitability of current and future investments, unconstrained firms do not have such cash adjustment motives. Therefore, a systematic pattern in cash policies for constrained firms is expected.

bolsters the precautionary demand for cash because internal resources are crucial to circumventing the adjustment costs inherent in R&D projects (Brown and Petersen, 2011). In the 21 st century, the demands for R&D of Japanese firms has grown steadily and financing of R&D projects has emerged as an important managerial issue. Although US firms can finance R&D projects by equity financing (Brown et al., 2009), this option has been very limited among Japanese firms. Moreover, lines of credits are not popular in Japan (Sasaki et al., 2014). Among Japanese firms, cash holdings play a dominant role in preparing for both future cash flow shortfalls and future investment opportunities. On the other hand, Pinkowitz and Williamson (2001) argue that banks that have monopoly power may extract rents from their client firms by demanding high cash reserves. Pinkowitz and Williamson (2001) present evidence consistent with the argument that high cash holdings in Japan can be attributed to the power of banks. The 2008 financial crisis may strengthen bank power and this increase in bank power may spur corporate cash holdings. It is worth noting that even after the deregulation of the bond markets in the 1990s, corporate bond issuances have been limited to firms with high corporate bond grades. We expect that firms without access to the bond markets voluntarily accumulate cash for the fear of losing access to bank loans. Although banks hardly require their client firms to accumulate bank deposits directly, client firms may

voluntarily increase bank deposits to maintain a good relationship with their main banks. Thus, we hypothesis that whereas firms with access to the bond markets can flexibly change their cash reserve policy based on precautionary demands such as financing of R&D projects, bank-dependent firms cannot do this to maintain good relationship with banks. 3. Research design and data 3.1. Research design We first compared the average cash holdings of bank-dependent firms and weakly-dependent firms (firms with access to the bond markets). We classified firms based on the bank loans-to-financial debts ratio (Bankloan). We classified firms whose average bank loan ratio has been 100% as bank-dependent and others as weaklydependent. We also investigated how R&D intensity affects cash holdings by estimating the average treatment effects with the propensity score matching strategy. Next, to examine how firm-bank relationship affects corporate cash saving behavior, we estimated cash flow sensitivities of cash following the specification of Almeida et al. (2004) as follows: ΔCashHoldingsi, t = αi +β1cfi, t +β2cf*rdi, t +β3qi, t +β4sizei, t + εi, t, (1),

where CashHoldings is measured as the sum of cash and cash equivalents; CF is defined as the operating cash flow divided by the total assets; RD is defined as R&D expenditure divided by the total assets, and its coefficient is our main concern; Q represents the ratio of the sum of the market value of equity and the book value of debt to total assets; and Size is the logarithm of total assets. In estimating equation (1), we ran regressions without firm fixed effects as well with firm fixed effects (FE) because firm fixed effects may absorb the important information regarding the corporate propensity to save and their inclusion may result in underestimations of CFSCs. We also submitted regressions without sector dummies because R&D intensities differ across industries rather than within an industry. As previously argued, internal cash flows are the main and steady sources for cash building for Japanese firms. We estimated equation (1) for both subsamples and compared the coefficients of CF for bank-dependent firms and weakly dependent firms. 3.2. Sample and descriptive statistics We collected data sampled from companies listed in the first section of the Tokyo Stock Exchange (TSE), and Nikkei Financial Quest. Our data sample comprises nonfinancial and non-utility firm-years for the period 2002 2011. We excluded financial

and utility companies as cash holdings of these firms may reflect the effects of regulation and do not necessarily reflect the extent of precautionary demands for cash. We trimmed outliers in key variables at the 1% level. Since we focused on the structure of financial debt to classify firms, we exclude firms without financial debts. 4 Table 2 presents the average cash holdings and other related variables for the full sample as well as the subsamples classified by the dependence on bank loans.. The mean of CashHoldings is 12.16%.The mean of ΔCashHoldings is 0.26% with a relatively large standard deviation of 3.79%. The average cash holding of firms that have access to the bond markets (weaklydependent) is 11.38%, whereas that of bank-dependent firms is 14.01%.. These figures are consistent with the results of Pinkowitz and Williamson (2001) and show that bankdependent firms have higher levels of cash regardless of the possibility that monitoring by main banks decrease symmetric information and agency costs. These high cash holdings among bank-dependent firms may be induced by the recovery of banks power after the 2008 financial crisis. Although it is unlikely that strong banks persuade firms to hold large cash balances, bank-dependent firms voluntarily save cash as deposits to 44 These firms have extensively higher cash flows as well as high levels of cash holdings. We think these financially unconstrained firms have precautionary motive for cash holdings. Investigating why these extremely cash-rich firms continue to hoard cash forms an interesting basis for future research.

maintain good relationships with their main banks. 4. Empirical results 4.1 Univariate analyses Table 3 compares the average of cash holdings between R&D-intensified firms and non-r&d-intensified firms for each subsample. We classified firms with higher (lower) R&D ratios than the 75 percentile (25 percentile) of R&D ratio within each subsample as R&D-intensified (non-r&d-intensified) firms. In both subsamples, the averages of cash holdings are larger for R&D-intensified firms. However, the difference in average cash holdings is larger for weakly-dependent firms than bank-dependent firms, suggesting that firms that rely less on bank loans can flexibly determine their cash policy, in response to precautionary demands for cash induced by R&D investment. On the contrary, bank-dependent firms have higher levels of cash holdings even if they do not have precautionary demands for cash. 4.2 Average treatment effects based on propensity score matching approach The simple comparison of cash holdings in Table 3 may be biased due to the possible self-selection bias that covariates associated with being R&D-intensified may

also affect cash holdings. Thus, we next compared the average cash holdings of R&Dintensified firms and other firms for the subsamples based on bank dependence, using a propensity matching estimate to cope with this possible self-selection bias. Specifically, we chose Tobin s Q, cash flow, and firm size as covariates in the probit estimation to calculate the propensity score for being R&D-intensified. These covariates have been used as regressors in R&D equations in previous studies (Himmelberg and Petersen, 1994, Brown and Petersen, 2009). We applied the nearest neighbor methodology in choosing control peers. Table 4 presents the average treatment effects for cash holding. Importantly, we found a significant difference of cash holdings only for firms that have access to the bond markets. This result again shows that bank-dependent firms have higher levels of cash holdings regardless of precautionary demands stemming from R&D investment. 4.3 Estimation of cash flow sensitivities of cash Next, we estimated the cash flow sensitivity of cash (CFSC) to further verify our hypotheses. Since the optimal levels of cash holdings may differ across firms, we complemented the findings thus far by investigating how bank dependence affected CFSCs, and the relationship between CFSCs and R&D. We measured corporate

propensity to save cash flows by estimating regressions of changes in cash holdings (ΔCashHoldings) on cash flows (CF) and other control variables. Following the specification of Almeida et al. (2004), we included Tobin s Q (Q) and firm size (Size) as control variables, which is defined as the logarithm of total assets. Table 5 presents the results. We observed that the coefficients of CF are larger for bank-dependent firms. For example, in the estimations with sector dummies (2nd and 5th columns, respectively), the coefficient of CF is 0.310 for weakly dependent firms and 0.382 for bank-dependent firms. These results indicate that bank-dependent firms have higher propensity to save internal cash flows. 4.4 Precautionary demands and cash flow sensitivity of cash So far, we have shown that bank-dependent firms have higher cash holdings and are more aggressive in saving cash flows, suggesting that bank-dependent firms prefer to accumulate cash to maintain good relationships with their main banks. However, this result may reflect the relationship between the extent of financing constraints and propensity to save cash, because bank-dependent firms can be considered more financially constrained and thus rely on bank loans. 5 To verify whether higher propensity 5 According to Sasaki et al. (2014), the percentage of CFOs that answered that their companies have access to the bond markets is lower for firms that rely on bank loans.

to save among bank-dependent firms comes from bank power or the extent of financing constraints, we estimated the regression wherein we inserted the intersection term between cash flow (CF) and R&D expenditure (RD), which is defined as R&D expenditure divided by total assets at the beginning of the fiscal year. If precautionary demands associated with R&D bolster corporate propensity to save cash flows, the coefficient of the intersection firm will be positive and statistically significant and especially so for bank-dependent firms. Contrary, if bank dependent firms attach importance to maintaining good relationship with banks in deciding cash saving policy, the coefficient of the intersection firms will not statistically significant for bank dependent firms. Table 6 presents the results. We reported the estimated results without fixed effects as firms use their cash holdings to finance R&D projects, and thus, the change in R&D of a firm in year t would not result in an immediate increase in cash holdings in year t from precautionary motives of saving cash. For firms that rely less on bank loans, the coefficients of the intersection term are all positive and statistically significant for two of three specifications. On the contrary, none of the coefficients of the intersection term is statistically significant for bank-dependent firms. We argue that while firms that rely Importantly, it shows a similar pattern for bank loans, suggesting that bank-dependent firms are likely to meet difficulty in raising capital from bank loans.

less on bank loans determine their corporate saving policy considering R&D demands, bank-dependent firms unanimously have higher propensity to save cash flows. These results suggest that bank-dependent firms cannot flexibly change their cash saving policy responding to precautionary demands for cash stemming from R&D investment. 4.5 Robustness tests 4.5.1 Investment variables and CFSCs We also submit the augmented regressions that include capital expenditure (Expenditures) and changes in net working capital (NWC) to control for the current demand for using cash holdings 6. Each of these is deflated by total assets. To cope with the possible endogeneity of these variables, we also run instrumental variable estimation in which Expenditures and NWC are instrumented by two lags of NWC, two lags of Expenditures, twice-lagged level of fixed capital, twice-lagged net working capital (the level of NWC), twice-lagged sales growth, and industry indicators. Table 7 presents the results. The coefficients of Expenditures and NWC are negative and statistically significant as expected. However, the inclusion of these variables do not change our main results. The coefficient of R&D*CF are positive and 6 Compared with the control variables in Almeida et al. (2004), we drop acquisitions because of unavailability of this variable. We also drop changes in short debts because Sargan tests reject the validity of the instruments when insert this variable as a regressor.

statistically significant for firms that have access to the bond markets, suggesting that precautionary demands for cash stemming from R&D investment is an important determinant for corporate propensity to save internal cash flows. Contrary, the coefficients of this cross-term are again not statistically significant for bank-dependent firms. These results again indicate that bank-dependent firms cannot flexibly determine cash saving policy. 4.5.2 Negative cash flows Next, we pay attention on the possibility that observations of negative cash flows may present a different pattern of CFSCs than observations of positive cash flows (Bao, Chang, and Zhang, 2012). We submit regressions that allow negative cash flow firm-years to have a differen CFSC than positive cash flow firm-years by including a cross-term between cash flow and a negative-indicator (1 for negative cash flow firm-years and 0 for others). Although we do not reproduce the results, our main results hold in this specification. The coefficients of cash flow are positive and statistically significant under all specifications and the coefficients of R&D*CF are statistically significant only for firms that have access to the bond markets.

4.5.3 Definition of cash holdings Finally, we check the robustness of our results to the definitions of changes in cash holdings. Specifically, we redefine changes in cash holdings that are free from changes in the market value of holding securities. As cash equivalents contain holding equities or bonds for investment purposes, changes in their market value may substantially impact changes in cash holdings. For this consideration, we use changes in cash holdings excluding gains / losses on the revaluation of marketable securities as the dependent variable and run the previously reported regression analysis. We do not reproduce the results here, but results are found to change very slightly and confirm our arguments that bank-dependent firms are aggressive in saving internal cash flows but cannot change saving policy responding precautionary demands for R&D investment. 5. Conclusion This paper investigates how Japanese firms accumulate cash holdings focusing on firm-bank relationship and precautionary demands for cash. We observed that bankdependent firms still have higher cash holdings than firms that have access to the bond markets. On the other hand, the differences in cash holdings between R&D-intensive firms and non-r&d-intensive firms are larger for firms that have access to the bond markets compared to bank-dependent firms. We also found that bank-dependent firms show larger cash flow sensitivity of cash, whereas the effect of R&D on cash flow

sensitivity is significant only for bank-dependent firms. These results suggest that firms that rely less on bank loans determine their cash policy based on precautionary demands for cash, whereas bank-dependent firms unanimously accumulate cash holdings to maintain good relations with banks. We argue that the growing demand for R&D and regaining of power by banks are the two main reasons explaining increase in cash holdings among Japanese firms following the financial crisis.

Reference Almeida, H., Campello, M., Weisbach, M.S., 2004. The cash flow sensitivity of cash. The Journal of Finance. 59, 1777-1804. Bao, D., Chan, K., C., and Zhang, W. 2012. Asymmetric cash flow sensitivity of cash holdings. Journal of Corporate Finance, 18: 690-700 Bates, T.W., Kahle, K.M., Stulz, R.M., 2009. Why do US firms hold so much more cash than they used to? The Journal of Finance. 64, 1985-2021. Brown, J.R., Fazzari, S.M., Petersen, B.C., 2009. Financing innovation and growth: Cash flow, external equity, and the 1990s R&D boom. The Journal of Finance. 64, 151-185. Brown, J., Petersen, B., 2009. Why has the investment-cash flow sensitivity declined so sharply? Rising R&D and equity market developments. Journal of Banking and Finance 33, 971 984. Brown, J.R., Petersen, B.C., 2011. Cash holdings and R&D smoothing. Journal of Corporate Finance. 17, 694-709. Hadlock, C.J., Pierce, J.R., 2010. New Evidence on Measuring Financial Constraints: Moving Beyond the KZ Index. Review of Financial Studies. 23, 1909-1940. Himmelberg, C.P., Petersen, B.C., 1994. R&D and internal finance: a panel study of small

firms in high-tech industries. Review of Economics and Statistics 76, 38 51. Iskandar-Datta, M.E., Jia, Y., 2012. Cross-country analysis of secular cash trends. Journal of Banking & Finance. 36, 898-912. Lins, K.V., Servaes, H., Tufano, P., 2010. What drives corporate liquidity? An international survey of cash holdings and lines of credit. Journal of Financial Economics. 98, 160-176. Opler, T., Pinkowitz, L., Stulz, R., Williamson, R., 1999. The determinants and implications of corporate cash holdings. Journal of Financial Economics. 52, 3-46. Pinkowitz, L., Williamson, R., 2001. Bank power and cash holdings: evidence from Japan. Review of Financial Studies. 14, 1059-1082. Pinkowitz, L., Stulz, R., Williamson, R., 2013. Is There a US High Cash Holdings Puzzle after the Financial Crisis? Fisher College of Business Working Paper. 07. Rajan, R.G., Zingales, L., 1995. What do we know about capital structure? Some evidence from international data. The journal of Finance. 50, 1421-1460. Sasaki, T, Sasaki, T., Xu, P., and Hanaeda, H., 2014. Cash holdings and liquidity management of Japanese firms. forcecoming in Proceeding of the 22th Nippon Finance Association Annual Meeting (in Japanese).

Subramaniam, V., Tang, T.T., Yue, H., Zhou, X., 2011. Firm structure and corporate cash holdings. Journal of Corporate Finance. 17, 759-773.

Average of Cash Holdings and Debt Ratios Figure 1 Development of corporate cash holdings and financial debts 0.3 0.2 0.1 Cash holding Debt ratio(bankloan) 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 This figure illustrates the amount of cash holdings and financial debts for the sample Japanese listed companies.

Average of Cash Holdings (by Bank dependence) Figure 2 Comparison of cash holdings for bank-dependent firms and bank-independent firms 0.3 Weakly dependent Dependent 0.2 0.1 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 This figure plots the average cash holding ratios for bank-dependent firms and bank-independent firms. The average cash holding ratio is defined as the sum of the cash and cash equivalents divided by the total assets. Firms that raise funding only by bank loans are classified as bank-dependent firms (dependent), while other firms with financial debts are classified as bank-independent firms (weakly dependent). This sample comprises nonfinancial and non-utility firm-years listed on the TSE for the period 2002 2011.

Table 1 Definitions of variables Variables Definition CashHoldings Cash and cash equivalents / total assets ΔCashHoldings ΔCash and cash equivalents / lagged total assets CF Cash flow / lagged total assets Q (Market value of equity + total debts) / assets Size Natural logarithm of lagged total assets Bankloan Bank loan / total debts R&D R&D expense / lagged total assets

Table 2 Descriptive statistics Panel A: Entire sample N mean sd min 25%ile 50%ile 75%ile max CashHoldings 10,491 0.1216 0.0820 0.0002 0.0605 0.1028 0.1630 0.6102 ΔCashHoldings 10,491 0.0026 0.0379-0.1326-0.0158 0.0007 0.0198 0.1735 CF 10,491 0.0580 0.0486-0.1100 0.0294 0.0572 0.0870 0.2311 Q 10,491 1.10 0.37 0.56 0.88 1.01 1.20 3.91 Size 10,491 11.64 1.20 9.03 10.79 11.47 12.36 14.91 Bankloan 10,491 0.8675 0.2109 0.0000 0.7967 1.0000 1.0000 1.0000 R&D 10,491 0.0169 0.0228 0.0000 0.0002 0.0082 0.0248 0.2846 Panel B: Weakly dependent N mean sd min 25%ile 50%ile 75%ile max CashHoldings 7,379 0.1138 0.0767 0.0002 0.0573 0.0972 0.1504 0.6102 ΔCashHoldings 7,379 0.0025 0.0368-0.1326-0.0152 0.0005 0.0188 0.1735 CF 7,379 0.0584 0.0469-0.1100 0.0312 0.0576 0.0863 0.2311 Q 7,379 1.11 0.35 0.56 0.90 1.02 1.20 3.78 Size 7,379 11.90 1.22 9.03 10.98 11.76 12.70 14.91 Bankloan 7,379 0.8116 0.2297 0.0000 0.7047 0.8924 1.0000 1.0000 R&D 7,379 0.0175 0.0233 0.0000 0.0003 0.0085 0.0257 0.2846 Panel C: Dependent N mean sd min 25%ile 50%ile 75%ile max CashHoldings 3,112 0.1401 0.0908 0.0003 0.0708 0.1225 0.1912 0.5466 ΔCashHoldings 3,112 0.0028 0.0405-0.1322-0.0176 0.0009 0.0229 0.1723 CF 3,112 0.0570 0.0525-0.1093 0.0241 0.0561 0.0900 0.2304 Q 3,112 1.0769 0.4173 0.5630 0.8383 0.9696 1.1744 3.9108 Size 3,112 11.04 0.89 9.03 10.41 11.00 11.56 14.52 Bankloan 3,112 1.0000 0.0000 1.0000 1.0000 1.0000 1.0000 1.0000 R&D 3,112 0.0156 0.0218 0.0000 0.0000 0.0074 0.0225 0.2155 This table reports the descriptive statistics of the variables used in the regression analyses for bankindependent firms (weakly dependent) and bank-dependent firms (dependent). Definitions of these variables are summarized in Table 1. This sample comprises nonfinancial and non-utility firm-years listed on the TSE for the period 2002 2011. We exclude zero-leverage firms, firms with debt exceeding assets, and outliers in key variables at the 1% level.

Table 3 Univariate analyses based on R&D intensity Panel A: Weakly dependent Low R&D High R&D Difference CashHoldings 0.1045 0.1224-0.0179 *** Panel B: Dependent Low R&D High R&D Difference CashHoldings 0.1369 0.1433-0.0064 ** This table presents the mean-comparison test by R&D intensity under a given bank dependence. Firms that raise funding only by bank loans are classified as bank-dependent firms (dependent), while other firms with financial debts are classified as bank-independent firms (weakly dependent). Firms with R&D expense over lagged total assets means that are less than the median are classified as Low R&D, while other firms with financial debts are classified as High R&D. Significance levels are reported at 10%(*), 5%(**), and 1%(***).

Table 4 Average treatment effects of being R&D intensified on cash holdings Panel A: Weakly dependent Low R&D High R&D Difference CashHoldings 0.1068 0.1224-0.0155 *** Panel B: Dependent Low R&D High R&D Difference CashHoldings 0.1374 0.1433-0.0059 This table reports the average treatment effects of being R&D intensified firms on cash holdings. We chose Tobin s Q, cash flow, and firm size as covariates in the probit estimation to calculate the propensity score for being R&D-intensified. We applied the nearest neighbor methodology in choosing control peers.

Table 5 Estimation of CFSCs for bank-dependent firms and bank-independent firms Weakly dependent Dependent OLS OLS PANEL FE OLS OLS PANEL FE CF 0.294 0.310 0.396 0.349 0.382 0.495 [22.06]*** [21.86]*** [21.38]*** [18.99]*** [18.92]*** [18.64]*** Q -0.005-0.006-0.011-0.008-0.007-0.012 [-2.49]** [-2.79]*** [-3.78]*** [-3.74]*** [-3.26]*** [-3.21]*** Size -0.001 0.000-0.021-0.001-0.001-0.012 [-1.93]* [-1.06] [-7.22]*** [-1.13] [-1.89]* [-2.24]** Constant -0.008-0.012 0.240-0.006 0.000 0.115 [-1.83]* [-2.52]** [6.76]*** [-0.69] [-0.02] [1.91]* Year dummy Yes Yes Yes Yes Yes Yes Industry dummy No Yes No No Yes No Observations 7,379 7,379 7,379 3,112 3,112 3,112 R-squared 0.15 0.16 0.19 0.21 0.23 0.25 Number of firms 924 471 This table presents regression estimates examining the CF sensitivities of annual changes in cash holdings for subsamples classified by bank dependence. Firms that raise funding only by bank loans are classified as bank-dependent firms (dependent), while other firms with financial debts are classified as bank-independent firms (weakly dependent). Clustered t-values within a firm are reported in brackets. Significance levels are reported at 10%(*), 5%(**), and 1%(***).

Table 6 Effects of R&D on CFSCs for bank-dependent firms and bank-independent firms Weakly dependent Dependent OLS OLS IV OLS OLS IV CF 0.288*** 0.288*** 0.259*** 0.342*** 0.339*** 0.250*** [19.27] [18.77] [17.66] [17.13] [16.51] [5.81] R&D*CF 0.729*** 1.110*** 1.143*** 0.287 0.776* 0.803 [2.94] [3.64] [3.79] [0.94] [1.74] [1.45] Q -0.001-0.001-0.002 0.000 0.001 0.004 [-0.43] [-0.51] [-1.10] [0.06] [0.27] [1.09] Size 0.000 0.000 0.000-0.00207*** -0.00215*** -0.00215** [-0.83] [-0.95] [-0.36] [-3.09] [-2.95] [-2.05] Expenditures -0.168*** -0.170*** -0.153*** -0.267*** -0.271*** -0.103 [-9.84] [-9.53] [-5.81] [-11.92] [-11.17] [-1.58] NWC -0.201*** -0.200*** -0.359*** -0.305*** -0.307*** -0.588*** [-18.83] [-18.63] [-10.77] [-15.01] [-14.99] [-4.40] Constant -0.0119*** -0.0115*** 0.004 0.010 0.009 0.019 [-2.98] [-2.77] [0.23] [1.13] [0.89] [1.33] Year dummy Yes Yes Yes Yes Yes Yes Industry dummy No Yes Yes No Yes Yes Observations 7,174 7,174 4,654 2,989 2,989 1,838 R-squared 0.25 0.26 0.24 0.37 0.38 0.27 Sargan statistic 2.516 12.67** [0.77] [0.03] This table presents regression estimates examining the CF sensitivities of annual changes in cash holdings for subsamples classified by bank dependence. R&D*CF are the cross-term of CF with Negative and RD, respectively. Firms that raise funding only by bank loans are classified as bankdependent firms (dependent), while other firms with financial debts are classified as bank-independent firms (weakly dependent). Clustered t-values within a firm are reported in brackets. Significance levels are reported at 10%(*), 5%(**), and 1%(***).

Table 7 Augmented regression results for the effects of R&D on CFSCs. Weakly dependent Dependent OLS OLS IV OLS OLS IV CF 0.288*** 0.288*** 0.259*** 0.342*** 0.339*** 0.250*** [19.27] [18.77] [17.66] [17.13] [16.51] [5.81] R&D*CF 0.729*** 1.110*** 1.143*** 0.287 0.776* 0.803 [2.94] [3.64] [3.79] [0.94] [1.74] [1.45] Q -0.001-0.001-0.002 0.000 0.001 0.004 [-0.43] [-0.51] [-1.10] [0.06] [0.27] [1.09] Size 0.000 0.000 0.000-0.00207*** -0.00215*** -0.00215** [-0.83] [-0.95] [-0.36] [-3.09] [-2.95] [-2.05] Expenditures -0.168*** -0.170*** -0.153*** -0.267*** -0.271*** -0.103 [-9.84] [-9.53] [-5.81] [-11.92] [-11.17] [-1.58] NWC -0.201*** -0.200*** -0.359*** -0.305*** -0.307*** -0.588*** [-18.83] [-18.63] [-10.77] [-15.01] [-14.99] [-4.40] Constant -0.0119*** -0.0115*** 0.004 0.010 0.009 0.019 [-2.98] [-2.77] [0.23] [1.13] [0.89] [1.33] Year dummy Yes Yes Yes Yes Yes Yes Industry dummy No Yes Yes No Yes Yes Observations 7,174 7,174 4,654 2,989 2,989 1,838 R-squared 0.25 0.26 0.24 0.37 0.38 0.27 Sargan statistic 2.516 12.67** [0.77] [0.03] This table presents augmented regression estimates examining the CF sensitivities of annual changes in cash holdings for subsamples classified by bank dependence. R&D*CF are the cross-term of CF with Negative and RD, respectively. Expenditures is defined as capital expenditures and NWC is changes in net working capital. Both are deflated by total assets at the beginning of the year. In IV regressions, Expenditures and NWC are instrumented by two lags of NWC, two lags of Expenditures, twice-lagged level of fixed capital, twice-lagged net working capital (the level of NWC), twice-lagged sales growth, and industry indicators. Firms that raise funding only by bank loans are classified as bank-dependent firms (dependent), while other firms with financial debts are classified as bankindependent firms (weakly dependent). Clustered t-values within a firm are reported in brackets. Significance levels are reported at 10%(*), 5%(**), and 1%(***).