The Effect of Financial Constraints, Investment Policy and Product Market Competition on the Value of Cash Holdings

Transcription

1 The Effect of Financial Constraints, Investment Policy and Product Market Competition on the Value of Cash Holdings Abstract This paper empirically investigates the value shareholders place on excess cash holdings and how shareholders valuation of cash holdings is associated with financial constraints, firm growth, cash flow uncertainty and product market competition for Australian firms from 1990 to Our results indicate that the marginal value of cash holdings to shareholders declines with larger cash holdings and higher leverage. However, firms that are more financially constrained, have higher growth rates and are facing higher uncertainty exhibit higher marginal value of cash holdings. These findings are consistent with the explanation that excess cash holdings are not necessarily detrimental to firm value. Firms with costly external financing and save more cash for current operating and future investing needs, find that the market values these cash hoarding policies favourably. Keywords: Financial Constraints, Cash Policy, Australian Firms. JEL Classification: G31, G32.

2 1. Introduction Are excess cash holdings good or bad? In the real world, excess cash holdings play a vital role as a cash buffer in corporate investment and financing decisions. Jensen (1986) holds the view that excess cash holdings are detrimental to shareholders value because managers will waste it through over-investment and value-destroying acquisition. In contrast, larger cash holdings provide firms flexibility in making investment decisions as it avoids the need to raise more costly external financing (Opler, Pinkowitz, Stulz, and Williamson (1999) and Mikkelson and Partch (2003)). A number of empirical studies such as Faulkender and Wang (2006), Mikkelson and Partch (2003) and Pinkowitz and Williamson (2004) investigate the value of corporate cash holdings and how excess cash holdings are related to stock returns in U.S markets. In Australia, Lee and Powell (2010) examine the value of excess cash holdings. Their findings support the agency cost argument of excess cash holdings by showing that firms with longer duration of excess cash holdings have lower marginal value of cash. However, Lee and Powell (2010) only show how the marginal value of cash is related to the length of excess holdings. They do not examine how shareholders value cash holdings as it relates to financial constraints, firm growth and cash flow uncertainty. Given the lack of Australian evidence, the focus of this paper is to examine how shareholders value excess cash holdings associated with financial constraints, firms growth opportunity, uncertainty in cash flows and product market competition. 1 1 The Australian market is different from the US. Their firms are generally smaller, there are more resource stocks 2

3 We first estimate the marginal value of cash holdings by following the methodology in Faulkender and Wang (2006). In particular, we use excess stock returns as the dependent variable and unexpected changes in cash holdings and firm characteristics that are related to cash as the explanatory variables. The excess stock returns are calculated using returns of the 25 Fama and French portfolios formed on market capitalization (size) and book-to-market as benchmark returns 2. Consistent with the results of Faulkender and Wang (2006), firms with larger cash holdings and higher leverage generate a lower marginal value of cash holdings. When we employ dividend-paying and book value of total assets to partition our sample according to the degree of financial constraints, we find more financially constrained firms have significantly higher marginal value of cash holdings. This indicates that investors value excess cash holdings of financially constrained firms more as these firms are less likely to get access to external capital when compared to unconstrained firms. In order to shed light on how the value of cash holdings is associated with firms investment decisions, uncertainty in cash flows and product market competition, we further partition our sample for the following variables: sales growth, average volatility in earnings and the Herfindahl index constructed using the market share of a firm s sales within the industry. We find that firms with higher growth rates and with higher level of uncertainty in their cash flows exhibit higher marginal value of cash holdings. and it has a much smaller and less liquid corporate bond market. 2 This approach follows Faulkender and Wang (2006). These benchmark portfolios formed on size and book-tomarket will account for common risk factors that affect stock returns. In contrast, Lee and Powell (2010) use excess returns adjusted by market returns instead of the 25 Fama and French portfolio returns. 3

4 However, product market competition has little impact on firms value of cash holdings. Our findings indicate that internal finance has clear cost advantages over external finance for companies with high growth potential and facing uncertain prospects. Costly external financing would force firms that are at a disadvantage to save more cash for current operating and future investing needs. Investors are aware of these cash hoarding policies and view them quite favourably. However, withinindustry product markets, competition seems to have little influence on a firms cash hoarding policy. Overall, our findings are mainly consistent with increased cash holdings being dependent on the firm s ability to access external capital as in Hennessy and Whited (2005). The remainder of this paper is organized as follows. Section 2 is a literature review which aims to discuss the existing studies surrounding corporate cash holdings and to link the cash-holding behaviour to various firm-specific factors in Australia. Section 3 discuss our empirical methodology and outlines our main hypotheses. The sample and summary statistics are described in Section 4. Section 5 presents our empirical results and robustness checks. Section 6 concludes the paper. 2. Literature Review Opler et al (1999) examine the determinants and implications of cash holdings and cash equivalents based on the data of 1048 publicly traded U.S. firms during the period between 1971 and Their findings show that the level of corporate cash holdings 4

5 is positively correlated with future investment opportunities, cash flow-to assets ratios, capital investment, industry volatility and investments in fixed assets, and is negatively correlated with firm size, leverage, and networking capital and dividend payments. In more recent papers, Faulkender and Wang (2006) extend this line of research by analysing the value that shareholders placed on the cash held by a firm. 3 They argue that the value of one additional dollar of cash reserves should decline with larger cash holdings, higher leverage and better access to capital markets. Their empirical findings support all of their above arguments including the idea that excess cash holdings are more valuable for shareholders in financially constrained firms. On the other hand, Pinkowitz and Williamson (2004) examine the value placed on a firm s cash holdings with different growth opportunities. They found that investors placed a higher value on firms cash holdings that had higher growth opportunities. In recent years, MacKay and Phillips (2005) investigate the effect of product market competition on a firm s financial choices. In particular, they examine leverage. They found that industry related factors and financial choices such as leverage are jointly determined. In addition, they found that firms in competitive industries used less financial leverage than those in less competitive industries. Fresard (2010) examines directly the role of cash holdings on a firms product market decision. He finds that cash holdings can be used to support competitive strategies against industry rivals. However, he does not directly investigate whether 3 The focus of their paper is not on how much cash is saved out of cash flow but on how the shareholders would value this by examining their excess stock returns. 5

6 firms in competitive industries will hold more cash holdings than those in less competitive industries and he does not examine whether investors place different value on excess cash holdings. Finally, in Australia, the main focus of the paper by Lee and Powell (2010) is to examine the determinants of the level of cash holdings in Australia. They find results similar to Opler et al (1999). In addition, they investigate the value of excess cash holdings in Australia. They show that firms with longer duration of excess cash holdings have lower marginal value of cash. However, they do not examine how shareholders value cash holdings as it relates to financial constraints, firm growth and cash flow uncertainty. 3. Empirical Methodology This section outlines the baseline empirical model involved in examining the value of cash holdings for Australian firms, followed by a discussion of the impact of financial constraints, firm growth opportunity, uncertainty in cash flows and product market competition on the marginal value of cash holdings. 3.1 Baseline Empirical Model The primary goal of this study is to investigate the value investors place on an extra dollar of cash held by firms, and how various factors that affect a firm external financing conditions would alter this value. Following Faulkender and Wang (2006), 6

7 we employ the following baseline model to regress excess stock returns over a fiscal year on the unexpected change in various firm-specific characteristics that affect cash positions in that fiscal year. This is represented by the following: r C E NA RND I D R B it, it, it, it, it, it, it, it, Mit, 1 Mit, 1 Mit, 1 Mit, 1 Mit, 1 Mit, 1 C NF C C C L * L * e it, it, 1 it, it, 1 it, it, 7 8 it, it, Mit, 1 Mit, 1 Mit, 1 Mit, 1 Mit, 1 (1) where represents the change in the variable X of firm i between fiscal year t and t-1. The dependent variable is the excess stock return where r i,t is the stock i s annualized return in fiscal year t and R B i,t is the annualized return of stock i s benchmark portfolio during fiscal year t. The benchmark portfolio is based on the whole sample of the AGSM Share Price & Price Relative (SPPR) database. We use the 25 Fama and French (1993) portfolios formed on market capitalization (size) and book-to-market equity ratios as our benchmark portfolios. The reason is that Fama and French (1993) found that size and book-to-market proxy for common risk factors. By controlling for these two common risk proxies, any relationship found between excess stock return and cash holdings would not be attributed to common risk factors. C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. RND t are capitalized Research and Development expenses, I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. Δ is notation for the change of variables from fiscal year t-1 to t. All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). As indicated by 7

8 Faulkender and Wang (2006), this methodology is a long-term event study that exploits the unexpected change of firm-specific factors to explain the abnormal returns. Specifically, the change in the value of cash reserves is the event while the entire fiscal year is defined as the event window. As in Faulkender and Wang (2006), we estimate the pooled OLS (ordinary least square) regression model given by equation (1) with year dummies. We allow residuals to be correlated within years by using the Huber-White variance/covariance matrix estimator to correct for potential heteroscadasticity. Petersen (2009) argues that both time-effect and firm-effect should be properly dealt with in a panel data set to mitigate estimation biases. In robustness checks, we also re-estimate our baseline model using the Fama-MacBeth method and the firm-fixed effect model. Our main results are qualitatively similar. For comparison with the results obtained in Faulkender and Wang (2006) and Lee and Powell (2010), we use the pooled OLS in our major regressions. 3.2 Financial Constraints and the Value of Cash In order to examine the impact of financial constraints on the marginal value of cash holdings, we need to classify our sample into financially constrained (FC) and nonfinancially constrained (NFC) firms 4. A firm is said to be FC if its cost of external 4 Some prior studies have used the KZ index (Kaplan and Zingales (1997) to classify firms as financially constrained (above-median KZ index) or unconstrained (below-median KZ index). The KZ index is constructed and based upon a small sample of 49 US firms. It may not be appropriate for Australian firms. Nevertheless and as a robustness check, we use the methodology from the Kaplan and Zingales (1997) paper to partition our sample into financially constrained and unconstrained firms. We find qualitatively similar results to those reported in this paper. 8

9 capital exceeds the cost of internal funds. However, this definition does not provide us with a clear-cut guidance on identifying constrained firms. Chang, Tan, Wong, and Zhang (2007) suggest that FC firms generally tend to have one or more of the following characteristics: small or unprofitable, high growth potential, high leverage and low debt capacity. Standard corporate finance theory suggests that smaller firms are more financially constrained than larger firms. As a result, we use an approach similar to Chang et al (2007). Our sample of firms is evenly divided into two groups according to their median book value of total assets. We define below-median firms as FC firms and above-median firms are defined as NFC firms 5. Smaller firms are more likely to be financially constrained than larger firms because they are typically young and less known to the market. Smaller firms are more likely to encounter information asymmetry and agency problems, which will make their external financing more expensive. Fazzari, Hubbard, and Petersen (1988) used the dividend payout approach to classify firms into financially constrained and non-financially constrained firms. They argue that due to information asymmetries in capital markets, FC firms have limited access to external finance. As a result, FC firms tend to retain most of their income. We follow Chang et al (2007) to partition our sample of firms into two groups according to their dividend payout ratios (as measured by dividend/ebit). Nondividend payers are firms that do not pay dividends and are more likely to be The results are not reported here but available upon request. 5 We also partition our sample of firms into tertiles (quintiles), and then compare the lowest tertile (quintile) with the highest tertile (quintile). Our results still hold. 9

10 financially constrained. Dividend payers are those paying dividends in a particular year and are viewed as unconstrained. We first follow Faulkender and Wang (2006) to test the following hypothesis. Hypothesis 1: The marginal value of cash holdings is negatively associated with the level of a firm s cash position and the level of a firm s leverage. Investors of financially constrained firms value excess cash holdings more than those of nonfinancially constrained firms. 3.3 Growth Rate, Uncertainty, Product Market Competition and the Value of Cash Firms with strong growth opportunities and investment needs are likely to hold more cash (Opler, Pinkowitz, Stulz, and Williamson (1999)). As a result, investors value cash holdings by firms with good growth opportunity at a premium to those with poor growth opportunity (Pinkowitz and Williamson (2004)). We use the growth rate of sales as a proxy for investment opportunity to partition our sample of firms into two groups: high sales group and low sales group 6. Hypothesis 2: Investors value excess cash holdings of firms with a high sales growth rate more than those firms with a low sales growth rate. 6 Using alternative proxies, such as growth in total assets, yield quite similar results. We do not use book-tomarket ratios as a proxy for growth opportunity as our dependent variable, excess stock return, has already adjusted for the book-to-market effect. 10

11 Firms facing high uncertainty in their cash flows are likely to hoard excess cash in fear of future cash short falls. Firms with high uncertain cash flows face higher external funding costs when compared to internally generated funds. As a result, firms with a high level of uncertainty are expected to rely more on internal funds and to save more cash. Opler et al (1999) find that firms with high risk cash flows generally hold relatively high levels of cash. We use standard deviation of earnings ratios (as measured by EBIT/total book assets) in the past five years to proxy for uncertainty in cash flows. Our sample is partitioned into high uncertainty and low uncertainty groups according to their standard deviation of earnings ratios. We propose the following hypothesis. Hypothesis 3: Investors value excess cash holdings of firms with high cash flow uncertainty more than those firms with low cash flow uncertainty. Finally, it has been argued in the literature by Phillips (1995), MacKay and Phillips (2005) and Fresard (2010) that intense product market competition would affect the firms financial choices. Firms in highly competitive industries are more likely to hoard more cash reserve as a buffer for their liquidity needs in the future. Therefore, firms with excess cash holdings are viewed positively by investors in the stock market. We use the Herfindahl-Hirschman Index (HHI) as a proxy for product market competition. The HHI is calculated by summing up the squares of the individual 11

12 market shares of sales for the firms in a specific industry. In this paper, we use the Centre for Research in Finance (CRIF) 26 Industry Classifications to calculate our HHIs. Our sample is partitioned into high industry competition (with low HHIs) and low industry competition (with high HHIs) groups according to their HHI at the industry level 7. Hypothesis 4: Investors value excess cash holdings of firms in highly competitive industries more than those of firms in industries with less competition. 4. Data 4.1. Sample Selection We start from a merged sample of firms listed on the Australian Stock Exchange over the period with accounting data and stock return data available from the Aspect Financial Database and the Australian Graduate School of Management Share Price and Price Relative (AGSM_SPPR) database. Firms are required to have no missing information for any of the following key variables including cash, earnings before interest and tax, total book assets, net interest expense, total common dividend paid, total debt, monthly stock returns and book value of equity from fiscal year t-1 to t. We limit our sample to the top 500 largest firms listed in the ASX according to their market capitalizations in these fiscal years. We also exclude firms in the financial and 7 We obtain similar results for our Tables 5, 6, 7 and 8 when comparing the lowest tertile (quintile) with the highest tertile (quintile) by using tertile- or quintile- partition. The results are not reported in the paper but available upon request from the corresponding author. 12

13 utilities sector (with CRIF industry code 18 Banks, 20 Insurance, 22 Real Estate Investment Trusts, and 26 Utilities) due to the relatively low physical capital investment of financials and the regulated nature of utilities. Our sample consists of 1,108 individual firms and 6,412 firm-year observations from 1990 to All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point 9. The dependent variable in the baseline regression is excess stock return. Monthly stock returns are required to calculate individual stock returns and benchmark portfolio returns. Following Faulkender and Wang (2006), we use the 25 Fama and French (1993) portfolios formed on market capitalization (size) and the ratio of book-tomarket equity. In particular, for each fiscal year, we sort firms into 25 size and bookto-market portfolios based on their market capitalization and book-to-market ratios. Then, the excess stock returns are calculated by subtracting annualized benchmark portfolio returns from annualized stock raw returns Summary Statistics The summary statistics of the main variables are reported in Table 1. Since all the explanatory variables except leverage ratio are scaled by the one-year lagged market value of equity, we could interpret our variables as changes in dollar-value. For 8 The minimum firms in a year is 221 (1990) and the maximum is 404 (2002, 2007). The average (median) over this period is approximately 356 (389) firms in a year. 9 Our results are qualitatively very similar when we truncate the distribution instead of winsorizing it. 13

14 example, the median firm has a -4.16% one-year excess stock return and a close to zero median change in cash holdings with a median cash holdings level of 5.47%. Table 1 indicates that on average, earnings before interest and tax (E t ), non-cash book assets (NA t ) and total common dividend paid (D t ) have been increasing over the sample period as both their mean and median values are positive. Table 1 Sample Summary Statistics ( ) Accounting data is obtained from the Aspect Financial Database and stock returns are obtained from the AGSM_SPPR Database for the fiscal years 1990 to Firms are required to have available information for and key variables needed in the study and to be the top 500 largest firms listed in the ASX according to their market capitalizations in these fiscal years. R i,t -RB i,t is the excess stock return, where R i,t is the annualized stock return of firm i in fiscal year t and RB i,t is stock i s annualized benchmark portfolio return in fiscal year t calculated on valueweighted returns of 25 portfolios, which are formed on size and book-to-market (5x5) as in Fama and French (1993). All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. Δ is notation for the change of variables from fiscal year t-1 to t. Variable Obs Mean Median Std. Dev. Min Max R i,t -RB i,t 6, Δ C t 6, C t-1 6, Δ E t 6, Δ NA t 6, Δ I t 6, Δ D t 6, L t 6, NF t 6, Panel A of Table 2 reports summary statistics of firms in our sample after we classify them into constrained and unconstrained categories. According to size and dividend payout, a financially constrained firm has on average a higher annualized excess returns, has a higher change in cash holdings, has a higher level of cash holdings, a higher leverage and a higher change in financing cash flow than does a financially unconstrained firm. Similarly, Panel B of Table 2 indicates that firms with 14

15 high sales growth and high level of cash flow uncertainty have high annualized excess returns, high changes in cash holdings and high changes in financing cash flow. However, firms in highly competitive industries have lower annualized excess returns, lower changes in cash holdings and slightly lower level of cash holdings 10. To summarize, summary statistics and univariate analyses indicate that financially constrained firms, firms with higher growth rates, firms with higher uncertainty, and firms with lower industry competition, exhibit higher annualized excess returns and hold more excess cash. Table 2 Financial Constraints, Sales Growth and Cash Flow Uncertainty Accounting data is obtained from the Aspect Financial Database and stock returns are obtained from the AGSM_SPPR Database for the fiscal years 1990 to Firms are required to have available information for and key variables needed in the study and to be the top 500 largest firms listed in the ASX according to their market capitalizations in these fiscal years. R i,t -RB i,t is the excess stock return, where R i,t is the annualized stock return of firm i in fiscal year t and RB i,t is stock i s annualized benchmark portfolio return in fiscal year t calculated on valueweighted returns of 25 portfolios, which are formed on size and book-to-market (5x5) as in Fama and French (1993). All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. Δ is notation for the change of variables from fiscal year t-1 to t. Firms are categorized as being financially constrained (FC) and unconstrained (NFC) according to their book value of assets and dividend payout ratio. T-statistics significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. Panel A Variable Dividend Paying FC NFC FC-NFC FC NFC FC-NFC R i,t -RB i,t *** *** Δ C t *** *** C t *** Δ E t *** * Δ NA t Δ I t *** *** Δ D t *** L t *** *** NF t *** *** Size 10 In the literature, there is no direct empirical evidence on whether firms in highly competitive industries would hold more or less cash. MacKay and Phillips (2005) only investigate the association between industry competition and leverage. In Panel B, the leverage ratios for highly-competitive industries are slightly lower than those for less competitive industries. This is consistent with MacKay and Phillips (2005). 15

16 Panel B Variable Sales Growth Cash Flow Uncertainty Industry Competition Low High High-Low Low High High-Low Low High High-Low R i,t -RB i,t *** *** *** Δ C t *** *** *** C t *** Δ E t *** *** Δ NA t *** *** Δ I t *** Δ D t *** *** L t *** *** NF t *** *** *** 5. Empirical Results 5.1. The Marginal Value of Cash Holdings Table 3 presents estimates from the baseline model for the entire sample using the pooled OLS regression. The initial coefficient estimations on the changes in cash holdings are statistically significant and positive at the 1% level. This suggests that an additional dollar of cash corresponds to AU$0.823 as valued by shareholders (Column (1)). The estimated coefficients on other control variables have the expected signs and are largely consistent with those reported in Faulkender and Wang (2006) and Lee and Powell (2010). In particular, the coefficients on changes in earnings, changes in dividends, level of cash holdings are positive and significant in all four columns. This confirms the results of Faulkender and Wang (2006). Further, the estimated coefficients are qualitatively similar when the following interaction terms: change in cash with the level of cash holdings (C t-1 * C t ), and the leverage ratio with the level of cash holdings (L t * C t ), are both included in the estimation. Based on the estimation in Column (4), the marginal value of cash to investors in the mean firm is 16

17 equal to AU$ (= AU$ ( * ) + ( * )) for the firm with 11.77% of cash to market value of equity and 19.39% of leverage ratio. This result is consistent with the first hypothesis that the marginal value of cash holdings is negatively associated with the level of a firm s cash position and the level of a firm s leverage ratio. Table 3 Regression Results for the Marginal Value of Cash Holdings This table presents the results of regressing the excess stock return R i,t -RB i,t on changes in firm characteristics including interaction terms between cash and leverage over the fiscal year over the fiscal year. All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. RND t are capitalized Research and Development expenses, I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point. T-statistics significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. VARIABLES (1) (2) (3) (4) Δ C t 0.823*** 0.729*** 0.726*** 1.095*** (9.58) (8.56) (8.50) (8.96) Δ E t 0.407*** 0.449*** 0.449*** 0.450*** (4.26) (4.51) (4.48) (4.86) Δ NA t 0.061*** (3.05) (1.62) (1.50) (1.54) Δ RND t (-0.50) (-0.44) (-0.56) Δ I t 1.270*** 1.351*** 1.059*** (2.96) (3.11) (2.63) Δ D t 0.671** 0.823** 0.738** (2.14) (2.56) (2.38) C t *** 0.466*** 0.466*** 0.416*** (7.08) (6.69) (6.63) (6.60) L t *** *** *** *** (-5.22) (-4.81) (-4.88) (-4.43) NF t 0.089* 0.094* (1.78) (1.88) (1.34) C t-1 * Δ C t *** (-2.98) L t * Δ C t ** (-2.53) Intercept ** ** (-0.90) (-1.15) (-2.26) (-2.47) Year dummy included No No Yes Yes Observations 6,412 6,412 6,412 6,412 R-squared

18 5.2 Financial Constraints and the Value of Cash Holdings Table 4 presents the estimated results for the association between financial constraints and the value of cash holdings. Columns (1) and (2) of Table 4 contain the regression results for the dividend-payout grouping. Non-dividend payers (Constrained) exhibit higher value of cash holdings than dividend payers (Unconstrained). Columns (3) and (4) report results obtained by using firm book value of assets as the measure of financial constraints. Similarly, small firms (Constrained) exhibit higher value of cash holdings than large firms (Unconstrained). On average, the marginal value of cash holdings for small firms is AU$ (= AU$ ( * 0.120) + ( * 0.135)) while the marginal value of cash holdings for large firms is AU$ (= AU$ ( * 0.116) + ( * 0.253)). The value difference for an additional dollar of cash holdings is AU$ This implies that investors place a significantly higher value on an extra dollar of cash holdings for financially-constrained firms since it is more difficult to access external financing for these firms. These results support our first hypothesis that investors of financially-constrained firms value excess cash holdings more than those of nonfinancially-constrained firms. In additional analysis, we perform a Chow test to see whether the estimated parameters differ statistically across these two financial constraint groups. The statistics in the last row of Table 4 show that the differences between the constrained and the non-constrained for dividend payout ratios are statistically significant with p- 18

19 values of less than 1%. The Chow test (with χ 2 statistic of ) shows that the two regressions are significantly different. Table 4 Regression Results for Financial Constraints This table presents the results of regressing the excess stock return R i,t -RB i,t on changes in firm characteristics for firms with and without financial constraints over the fiscal year. The baseline model is pooled OLS model controlling for time effect. All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. RND t are capitalized Research and Development expenses, I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point. T-statistics significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. Constrained Unconstrained Constrained Unconstrained VARIABLES Non-Dividend Payers Dividend Payers Small Firms Large Firms Δ C t 1.108*** 0.520*** 1.213*** 0.701*** (10.20) (6.78) (12.71) (8.68) Δ E t 0.315*** 0.992*** 0.465*** 0.455*** (3.84) (15.83) (6.41) (8.10) Δ NA t *** *** (-0.01) (6.47) (1.49) (5.32) Δ RND t (-0.40) (-0.59) (-0.70) (-0.47) Δ I t *** *** (1.02) (7.27) (0.25) (4.98) Δ D t *** 1.473*** (-1.11) (4.58) (2.95) (0.52) C t *** 0.213*** 0.772*** 0.198*** (6.28) (4.68) (9.44) (5.99) L t *** ** * *** (-2.86) (-2.18) (-1.80) (-4.60) NF t 0.203*** *** 0.109** *** (3.96) (-6.66) (2.49) (-2.59) C t-1 * Δ C t *** *** *** (-5.77) (-0.01) (-6.23) (-2.79) L t * Δ C t ** *** (-2.39) (-0.07) (-1.05) (-3.64) Intercept * *** (-1.94) (-1.19) (-4.36) (1.28) Observations 2,081 4,331 3,211 3,201 R-squared Difference between Constrained and Unconstrained (χ 2 of Chow Test) *** 92.16*** 19

20 5.3 Growth Opportunity, Uncertainty, Product Market Competition To investigate whether growth opportunity, uncertainty in cash flows and product market competition would affect the value of cash holdings in a similar way as financial constraints, we use sales growth rate (proxy for growth opportunity), standard deviation of earnings (proxy for cash flow uncertainty) and the Herfindahl-Hirschman Index (proxy for product market competition) to partition our sample. Table 5 reports results for the association between growth opportunity and the value of cash holdings. Low growth firms have lower value of cash holdings than high growth firms. On average, the marginal value of cash holdings for low growth firms is AU$0.675 (= AU$ ( * 0.120) + ( * 0.200)) while the marginal value of cash holdings for high growth firms is AU$0.979 (= AU$ ( * 0.116) + ( * 0.187)). The dollar value difference is AU$ per additional dollar of cash holdings. This confirms our hypothesis two that investors value excess cash holdings of firms with high sales growth more than those of firms with low sales growth rate. The results for the association between cash flow uncertainty and the value of cash holdings are presented in Table 6. It can been seen from the table that firms with high uncertainty in cash flows exhibit higher value of cash holdings than firms with low uncertainty. On average, the marginal value of cash holdings for high uncertainty firms is AU$1.014 (= AU$ ( * 0.127) + ( * 0.165)) while the marginal value of cash holdings for low uncertainty firms is AU$0.626 (= AU$

21 ( * 0.108) + ( * 0.222)). The Chow test (with χ 2 statistic of 67.89) shows that the two group regressions are significantly different. The results support our hypothesis three that firms with high cash flow uncertainty have higher marginal value in excess cash holdings than firms with low cash flow uncertainty. Table 5 Regression Results for Sales Growth This table presents the results of regressing the excess stock return R i,t -RB i,t on changes in firm characteristics for firms with different sales growth rate over the fiscal year. The baseline model is pooled OLS model controlling for time effect. All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. RND t are capitalized Research and Development expenses, I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point. T-statistics significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. VARIABLES Low Sales Growth High Sales Growth Δ C t 0.828*** 1.233*** (9.75) (14.28) Δ E t 0.472*** 0.423*** (7.47) (5.85) Δ NA t 0.073*** (3.91) (0.96) Δ RND t ** (-2.39) (0.41) Δ I t 1.485*** 0.945** (4.94) (2.10) Δ D t 1.034*** (3.17) (0.95) C t *** 0.598*** (7.02) (8.28) L t ** *** (-2.51) (-3.58) NF t (1.27) (1.45) C t-1 * Δ C t *** *** (-2.70) (-6.54) L t * Δ C t ** *** (-2.57) (-2.70) Intercept *** (-1.33) (-2.63) Observations 3,211 3,201 R-squared Difference between Low and High Sales Growth Firms (χ 2 of Chow Test) 42.90*** 21

22 Table 6 Regression Results for Cash Flow Uncertainty This table presents the results of regressing the excess stock return R i,t -RB i,t on changes in firm characteristics for firms with different cash flow uncertainty over the fiscal year. The baseline model is pooled OLS model controlling for time effect. All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. RND t are capitalized Research and Development expenses, I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point. T-statistics significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. VARIABLES Low Uncertainty High Uncertainty Δ C t 0.799*** 1.207*** (10.92) (13.09) Δ E t 0.183* 0.471*** (1.95) (7.23) Δ NA t ** (1.58) (2.01) Δ RND t *** (-2.98) (0.28) Δ I t 1.145*** 0.920** (4.17) (2.14) Δ D t *** (0.96) (2.79) C t *** 0.705*** (6.05) (9.23) L t *** *** (-4.30) (-2.66) NF t 0.076*** (2.59) (0.89) C t-1 * Δ C t *** *** (-5.87) (-3.38) L t * Δ C t *** ** (-2.86) (-2.43) Intercept ** ** (-1.97) (-2.31) Observations 3,211 3,201 R-squared Difference between Low and High Cash Flow Uncertainty Firms (χ 2 of Chow Test) 67.89*** Table 7 reports results for the association between product market competition and the value of cash holdings. In contrast with the findings in sales growth and cash flow uncertainty, product market competition seems have little influence on the value of 22

23 excess cash holdings. The Chow test (with χ 2 statistic of 26.07) also indicates that the two industry-competition group regressions are not statistically and significantly different. Table 7 Regression Results for Industry Competition This table presents the results of regressing the excess stock return R i,t -RB i,t on changes in firm characteristics for firms with different industry competition provide by the Herfindahl index over the fiscal year. The baseline model is pooled OLS model controlling for time effect. All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. RND t are capitalized Research and Development expenses, I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point. T-statistics significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. VARIABLES High Industry Competition Low Industry Competition Δ C t 1.167*** 0.986*** (14.73) (10.85) Δ E t 0.356*** 0.564*** (5.96) (7.41) Δ NA t 0.079*** (4.61) (0.60) Δ RND t (0.51) (-1.21) Δ I t 0.944*** 1.424*** (3.03) (3.23) Δ D t 0.816** (2.34) (1.59) C t *** 0.416*** (7.90) (6.39) L t *** *** (-3.26) (-3.34) NF t *** (-0.94) (3.61) C t-1 * Δ C t *** *** (-4.46) (-4.70) L t * Δ C t *** ** (-4.72) (-2.03) Intercept *** (-3.16) (-1.04) Observations 3,211 3,201 R-squared Difference between Low and High Industry Competition Firms (χ 2 of Chow Test) However, when we calculate the difference in the value of excess cash holdings, the marginal value of cash holdings for firms in highly competitive industries (AU$

24 = AU$ ( * 0.114) + ( * 0.191)) is slightly higher than the marginal value of cash holdings for firms in industries with less competition (AU$0.804 = AU$ ( * 0.121) + ( * 0.197)). This evidence of value differences partly supports our hypothesis four. 5.4 Robustness Checks Our results reported in the previous sections are obtained using the pooled OLS regressions corrected for heteroskedasticity. Peterson (2009) shows that in the presence of time effect and firm effect, the pooled OLS regressions in a panel data set will be biased. Traditionally, the Fama and MacBeth (1973) approach has been undertaken to account for time effect. However, the Fama-MacBeth approach is only effective to produce unbiased result when there is only time effect in place. It is possible the effect of change in cash on excess stock return is caused by some unobserved firm-specific factor(s). The previous studies advocate the firm fixed-effect model to control for unobservable time-invariant firm heterogeneity (for example, Himmerlberg, Hubbard, and Palia (1999)). We employ both approaches to check the robustness of our findings using the pooled OLS regressions in Table 3. The results for the Fama-MacBeth regressions and the firm fixed-effect regressions are reported in Table 8. Consistent with the results estimated using the pooled OLS approach, the main estimated coefficients have the same predicted signs and magnitudes. This 24

25 suggests that using alternative estimation approaches would generate qualitatively similar results. Table 8 Robustness Checks This table presents the results of regressing the excess stock return R i,t -RB i,t on changes in firm characteristics including interaction terms between cash and leverage over the fiscal year over the fiscal year. All variables except L t (Leverage) and excess stock return are deflated by the lagged market value of equity (M t 1 ). C t is cash including short term deposits, E t is earnings before interest and tax (EBIT), and NA t is total book assets minus C t. RND t are capitalized Research and Development expenses, I t is net interest expense, D t is total common dividend paid, L t is total debt divided by total book assets, and NF t is the net changes in total financing cash flow. All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point. T-statistics significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. VARIABLES Fama-MacBeth Fama-MacBeth Firm Fixed Effects Firm Fixed Effects Δ C t 0.718*** 0.917*** 0.807*** 1.074*** (6.40) (5.40) (15.78) (16.15) Δ E t 0.54*** 0.56*** 0.477*** 0.496*** (4.68) (5.24) (8.97) (9.32) Δ NA t 0.068** 0.064** 0.046*** 0.042*** (2.24) (2.25) (3.06) (2.81) Δ RND t * ** (-0.97) (-1.02) (-1.92) (-2.08) Δ I t 1.600*** 1.591*** 1.178*** 0.938*** (3.33) (3.38) (4.11) (3.25) Δ D t 0.811* 0.834* 0.861*** 0.815*** (2.04) (2.13) (2.97) (2.82) C t *** 0.472*** 0.913*** 0.847*** (6.89) (6.78) (15.49) (14.11) L t ** ** *** *** (-2.36) (-2.35) (-5.24) (-4.72) NF t ** 0.063** (1.26) (1.13) (2.54) (2.02) C t-1 * Δ C t * (-1.26) (-1.81) L t * Δ C t *** (-0.89) (-5.20) Intercept * ** *** *** (-1.91) (-2.36) (-3.84) (-3.98) Year dummy included No No Yes Yes Observations 6,412 6,412 6,412 6,412 R-squared

26 In equation (1), the dependent variable is defined as excess stock returns whilst the other variables are scaled by the lagged market value of equity. This allows us to interpret our findings as to how investors value excess cash holdings in dollar terms. In the empirical literature, an alternative approach is to scale by the total book value of assets. To check the robustness of our findings, we next employ a value regression as outlined in Fama and French (1998). We use changes in market value over the total book value of assets as the dependent variable and variables likely to affect the firms future cash flows (scaled by total book value of assets) as the explanatory variables. Specifically, we estimate the following regression for each measure of our financial constraint proxies, sales growth, cash flow volatility and industry competition: E de de da da RND ( M A )/ A a b b b b b b it, it, it, 2 it, it, 2 it, it, it, it, Ait, Ait, Ait, Ait, Ait, Ait, drnd drnd I di di b b b b b it, it, 2 it, it, it, Ait, Ait, Ait, Ait, Ait, Dit, ddit, ddit, 2 dmit, 2 Cit, it, Ait, Ait, Ait, Ait, Ait, + b b b b b (2) where d X t represents the two-year change in the variable X, X t - X t-2. M t is total market value of assets, and A t is total book value of assets. E t is earnings before interest and tax (EBIT), RND t are capitalized Research and Development expenses, I t is net interest expense, D t is the total common dividend paid, C t is cash including short term deposits. All variables are deflated by the total book value of assets (A t ) as in Fama and French (1998). For brevity, we only report the estimated coefficients on C t (b 16 ) for the whole sample and for each measure of our financial constraints proxies such as sales growth, cash flow volatility and industry competition. Consistent with the results 26

27 estimated using equation (1), the main estimated coefficients have the same predicted signs and magnitudes. This further confirms our main hypotheses that more financially constrained firms, firms with higher growth rates, with higher product market competition and with higher level of uncertainty in their cash flows exhibit signifcantly higher market value. Table 9 Regression Results for the Market Value of Assets This table presents the results of regressing the two-year change in market value of assets on changes in firm characteristics scaled by total book value of assets. Firms are categorized as being financially constrained (FC) and unconstrained (NFC) according to their book value of assets and dividend payout ratio. ΔCt column reports the estimated coefficients on ΔCt in Equation (2). All explanatory variables are deflated by the total book value of assets. All variables have been winsorized at the 1st and 99th percentiles. This approach reduces the impact of extreme observations by assigning the cut-off value to values beyond the cut-off point. T-statistics (reported in parentheses) significant at the 10%, 5% and 1% level are marked with *, ** and *** respectively. Δ C t Observations R-squared Whole sample 1.533*** (5.47) 4, Non-Dividend Payers 0.871** (2.47) 1, Dividend Payers 0.771** (2.47) 3, Small firms 1.260*** (3.89) 2, Large firms 0.934** (2.39) 2, High sales growth 1.735*** (4.87) 2, Low sales growth 0.973** (2.15) 2, High cash flow volatility 1.252*** (2.80) 2, Low cash flow volatility 1.169*** (3.60) 2, High industry competition 2.190*** (4.87) 2, Low industry competition 0.921*** (2.65) 2, Conclusions In this paper, we investigate whether firms with larger cash holdings and higher leverage would generate lower marginal value of cash holdings, and whether financial 27