The Determinants of Corporate Cash Holdings: A Comparison Between Brazilian and US Firms

Transcription

1 The Determinants of Corporate Cash Holdings: A Comparison Between Brazilian and US Firms Senichiro Koshio Escola de Administração de Empresas de São Paulo, Fundação Getúlio Vargas sekoshio@gvmail.br Joanília Neide de Sales Cia Escola de Administração de Empresas de São Paulo, Fundação Getúlio Vargas joanilia@uol.com.br

2 The Determinants of Corporate Cash Holdings: a Comparison between Brazilian and US Firms Abstract: This study investigates the determinants of cash holdings (defined as cash and bank deposs plus short term financial investments) by publicly traded non-financial firms in Brazil in comparison to those of US firms from 1995 to The results indicate that many of the determinants and the tradeoffs associated wh them in Brazil are different from those in the US: whereas the most significant determinants are the operating income, inventories, account payables and debts in Brazil, they are the account receivables and account payables in the US; the operating income affects posively the cash holdings in Brazil, whereas s effect is negative in the US; and, the market-to-book ratio is a determinant in the US, but not in Brazil. In addion, there are evidences of the structural changes in the determinants of corporate cash holdings over time in two countries. The findings also indicate that the determinants cannot be explained totally neher by the tradeoff theory nor by the buffer stock liquidy theory in both countries; however, they may be explained in a combined setting of these two theories. I. Introduction Why firms hold cash (cash bank deposs short term financial investments), sometimes in very large amounts? The investigations to answer this apparently simple question gained dynamism just few years ago, since Opler, Pinkowz, Stulz and Williamson (hereafter OPSW, 1999) presented the evidences that the cash holdings are based on static tradeoffs between their benefs and costs in the US. And, together wh some studies before them, the lerature on the subject has been enriched progressively by researches focusing on the effects of firms internal or external factors to explain the existence of cash holdings: costs of external financing [Kim et al. (1998)]; value destroying acquision behavior of cash-rich firms [Harford (1999)]; firm-bank relationship in Japan [Pinkowz and Williamson (1999)]; long-term debt in UK and Belgium [Anderson (2002)]; high inflation and economic instabily in Brazil [Falls and Natke (1996) and Natke (2001)]; legal and ownership structure in UK, Japan, France and Germany [Ozkan and Ozkan (2002); Guney, Ozkan and Ozkan (2003)]; countries instutional qualies (country risk and investor protection) [Dtmar, Smh and Servaes (2002); Pinkowz, Stulz and Williamson (2004)]; and, firms growth opportunies [Pinkowz and Williamson (2002); Pinkowz, Stulz and Williamson (2003)]. In fact, the tradeoff theory, consistent wh the objective of shareholder value maximization, can be considered as the main stream theory on the corporate cash holdings, likewise many other topics in Economics and Finance; however, there is also an alternative to this theory: the buffer stock liquidy theory, presented in Bruinshoofd and Kool (hereafter BK, 2002) in a reconciled model setting, under which the cash holdings follow the tradeoff theory in the long run, but the buffer stock liquidy theory in the short run, wh empirical evidences in the Netherlands. In Brazil, Koshio and Cia (hereafter KC, 2003) analyzed the determinants of corporate cash holdings for the period and found evidences that many determinants in Brazil were different from those in the US, reported by OPSW (1999) for the period, and that the determinants in Brazil followed more the buffer stock liquidy theory 1 than the tradeoff theory, whereas the determinants in the US followed more the latter than the former. 1

3 Nevertheless, the differences in the determinants in Brazil in relation to the US found by KC (2003) might also have been due to the application of different regression model specification as well as to the differences in the time periods of study. The objective of this paper, therefore, is to compare the determinants of corporate cash holdings by Brazilian and US nonfinancial firms, applying the same regression models to the same time period, and verify more rigorously the consistency between the determinants of cash holdings wh the tradeoff and the buffer stock liquidy theories in two countries. Our results indicate that many of the determinants and the tradeoffs associated wh them in Brazil are indeed different from those in the US: whereas the most significant determinants are the operating income, inventories, account payables and debts in Brazil, they are the account receivables and account payables in the US; the operating income affects posively the cash holdings in Brazil, whereas s effect is negative in the US; and, the market-to-book ratio is a determinant in the US, but not in Brazil. In addion, there are evidences of the structural changes in the determinants of corporate cash holdings over time in two countries. The findings indicate also that the determinants cannot be explained totally neher by the tradeoff theory nor by the buffer stock liquidy theory in both countries; however, they may be explained in a combined setting of these two theories, which is consistent wh BK (2002). The remaining of the paper is structured as follows: is discussed the static tradeoff theory and the buffer stock liquidy theory on corporate cash holdings in the next section, the empirical methodology in the section III, the utilized data in the section IV, the results of the tests and their analyses in the section V, and the conclusion in the last section. II. Static Tradeoff Theory and Buffer Stock Liquidy Theory on Cash Holdings Under the hypothetical perfect financial markets, the cash holdings are irrelevant for firm values, since firms are able to get funds wh no friction whenever need them. Under the real condions, however, there are several financial market imperfections which make them relevant; and, according to the static tradeoff theory, the cash holdings are actively managed in order to maximize their net benefs, e.g., the benefs minus costs, at each point in time. The main benefs of cash holdings, as stated by Keynes (1936), are the reduction of transaction costs, the cash buffer for precaution and the cash allowance for speculation. The practice of arbrage, taking advantage of differences in interest rates between foreign and domestic financial markets to make free risk profs, is another benef valid specifically in Brazil, as described in KC (2003). On the other hand, the main costs of cash holdings are the opportuny cost of capal and the liquidy premium. By equaling the marginal costs to the marginal benefs of holding cash, the decision leads to the optimal levels of cash holdings, consistent wh the shareholder value maximization objective. And, in this sense, the higher the benefs and lower the costs, the higher the cash holdings by firms. OPSW (1999), KC (2003) and other researchers have investigated the determinants of corporate cash holdings regressing them on the potential explanatory variables, many of which are based on the firms accounting variables, and have interpreted them considering the predictions of the tradeoff theory. In our opinion, however, these predictions should be revised, incorporating the benefs and costs the firms may have in a broader perspective: (a) Firm size: negative coefficient, e.g., the greater the size, the lower the cash holdings. Larger firms tend to have easier access to the financial markets, which provide them wh funds when required and, therefore, they may keep lower cash holdings. 2 2

4 (b) Operating income (operating return over assets): negative coefficient, e.g., the greater the operating income, the lower the cash holdings. The firms wh greater operating income have lower needs of cash buffers to cover eventual losses. 3 (c) Variation in net fixed assets: posive coefficient, e.g., the greater the variation in net fixed assets, the higher the cash holdings. A posive variation in net fixed assets, or new investments in fixed assets, which are risky assets in general, imply in higher cash holdings for the transaction (to reduce the transaction costs paying at sight), precaution (to cover eventual losses wh these assets) and speculation (to take advantage of timely investment opportunies) motives. A negative variation in net fixed assets imply in sales of fixed assets or depreciation higher than new investments, which is a source of funds; and, this cash inflow allows firms to keep less cash holdings. In addion, fixed assets are not good substutes to the cash holdings, as they are the assets wh low liquidy in general, not allowing firms to hold lower cash. (d) Variation in capal investments (investments in subsidiaries and other companies, including company acquision): posive coefficient, e.g., the greater the capal investments, the higher the cash holdings. Similarly to the net fixed assets, a posive variation in capal investments implies in higher cash holdings for the transaction, speculation and precaution motives; and, a negative variation in capal investments, which is a source of funds implies in cash inflow, which lets firms to keep less cash holdings. There is also no substution effect wh the capal investments due to s low liquidy. On these four already mentioned variables, there is only one coefficient sign predicted; however, there are no clear-cut predictions for other variables: the signs may be posive and negative: (e) Account receivables: greater receivables imply in lower cash holdings, taking into account the possibily of trading them in the financial markets and converting them into cash, whenever necessary; however, may also be that greater receivables imply in higher cash holdings, considering the risk of non-payment of these receivables and the consequent needs of reserves in cash to cover the resulting losses. (f) Inventories: greater inventories imply in lower cash holdings, taking into account the possibily of selling them out in the product markets and converting them into cash or of using them as pledge to get loans in the financial markets. Inventories also work as a reserve of value in an inflationary economy, avoiding to incur in losses wh other current assets not indexed to the inflation, or a mean to hedge against the risk of default of treasury securies or against confiscation of monetary assets in economically and polically instable countries such as Brazil, implying in direct conversion of cash to inventories in these cases. However, may also be that greater inventories imply in higher cash holdings, considering the risk of decrease in their values, for instance, by obsolescence or decrease in qualy and, therefore, reserves in cash are required to cover the resulting losses. (g) Account payables: greater account payables imply in higher cash holdings, considering the risks of incurring in addional costs to negotiate the payment condions wh suppliers (transaction motive) and of having to make payments in larger amounts or in shorter time periods than the originally expected (precaution motive). Notwhstanding, greater account payables may also imply in lower cash holdings, taking into account the possibily of delaying the payment to the suppliers when is possible or necessary and, therefore, making easier the management of the cash shortage suation. (h) Current debt: a higher debt-to-asset ratio implies in higher cash holdings, taking into consideration the transaction costs to obtain new loans or the precaution for the case of not being able to roll-over the loans in the short-term. There is also the arbrage motive to have debt and cash at the same time, which is applicable particularly in Brazil. However, a 3

5 higher debt-to-asset ratio may also imply in lower cash holdings when the loans can be easily managed, for instance, when they are easy to be rolled over or their repayments may be delayed, which is a suation of very low transaction costs, legal or instutional fragily or high bargaining power of the debtor in relation to the credor. (i) Long-term debt: the basic reasoning is the same as the current debt. The difference in this case is that there is no reasoning related to debt roll-over or repayment as their due dates occur in long-terms. Its posive relation wh the cash holdings is basically due to their high transaction costs and consequent high volume and discrete event natures. The relation wh the cash holdings may also be negative, however, taking into account the effect of better cred and banking relationship the firms wh long-term financing can have in the financial markets, letting the firms have easier access to the short-term emergency loans, such as hot moneys, and, therefore, hold lower cash. (j) Dividends: the firms which pay out dividends tend to have lower cash holdings, as they can decide to postpone or even not to pay out the dividends when they are short in cash. However, on the other hand, the firms which are obliged to pay out dividends by law or company bylaw, tend to have higher cash holdings to meet the payment needs. (k) Market-to-Book ratio: as a proxy for the growth opportunies, firms wh high market-to-book ratio have higher growth opportunies and, therefore, tend to hold higher cash for the speculation motive. However, may also be considered as a proxy for accessibily to the financial markets; and, in this case, as firms wh high market-to-book ratio may attract more capal in the financial markets, they can hold lower cash. An alternative theory on corporate cash holdings we consider is the buffer stock liquidy theory of BK (2002), under which the firms do not make any active decision related to the cash holdings themselves. Instead, they leave the cash holdings as a consequence of other decisions made by the firms, which imply in cash inflow when the decision is related to a source of funds or in cash outflow when is related to a use of funds. In this sense, cash holdings work as a mere buffer stock between cash inflow and cash outflow, process in which any shortage in this buffer stock is met by the funds provided following some financing decision creria. 4 The predictions under the buffer stock liquidy theory are: (a) Firm size: posive coefficient. The larger the firm, the larger the cash holdings, since a large firm size is a result of past net posive cash inflows, which were accumulated in greater amount whin the firm in comparison to smaller firms in the past. (b) Operating income: posive coefficient. The greater the operating income, the higher the cash holdings, since more profable firms tend to generate and accumulate more cash. (c) Variation in net fixed assets, variation in capal investments, account receivables, inventories and dividends: negative coefficients. Their increases are uses of cash, whereas their decreases are sources of funds. (d) Account payables, current debt and long-term debt: posive coefficients. Their increases are sources of funds, whereas their decreases are uses of funds. (e) Market-to-Book ratio: there is no effect on the cash holdings. It is not related to the source or use of funds under the buffer stock liquidy theory. In sum, the firm size, operating income, variation in net fixed assets and variation in capal investments have the oppose predictions in terms of the signs of the regression coefficients between the tradeoff theory and the buffer stock liquidy theory; nevertheless, other variables have predictions that can be both posive and negative under the tradeoff theory, and one of these two signs is the same as the predicted by the buffer stock liquidy theory. Although this fact makes not possible the clear identification of the results of 4

6 regressions between the predictions of two theories in some cases, we can still examine the extent to which the regression results conform to these two theories and analyze the firms behaviors in practice. III. Methodology In order to compare the determinants of corporate cash holdings in Brazil and in the US, we first test the variables mentioned in the previous section as the determinants of corporate cash holdings in two countries, applying the same regression models for the same time period. Next, we test whether the coefficients are statistically different between two countries. Then, we test whether there are statistically significant changes in the coefficients over time in each country. And, finally, we interpret the regression coefficients wh the aim of verifying the conformy of the determinants wh the tradeoff theory or the buffer stock liquidy theory as well as analyzing the differences in the applied tradeoffs between the Brazilian and the US firms. A) The determinants of corporate cash holdings in Brazil and in the US In order to examine the determinants, we apply the liner regressions on panel data formed wh firms accounting and stock market values, regressing the cash holdings (the dependent variable) on the potential determinants (the explanatory variables). The signs of the coefficients obtained by regressions indicate the effects of the explanatory variables on the cash holdings in each country. The variables and their definions are listed in the table 1. Table 1. Variable Definions for Regression Analyses Variable lcash lsize opinc vnetfa vcapinv arec invent apay crdebt ltdebt ddivid mtb Definion ln [1 Cash / (Total assets Cash)] ln [Total assets in Dec/2002 currency] Operating income / (Total assets Cash) Variation in net fixed assets / (Total assets Cash) Variation in capal investments / (Total assets Cash) Account receivables from trade / (Total assets Cash) Inventories / (Total assets Cash) Account payables / (Total assets Cash) Current debt / (Total assets Cash) Long-term debt / (Total assets Cash) One if paid out or zero if did not pay out dividends (Total assets - book equy value market equy value) / (Total assets) Notes: All balance sheet variables use the end of year figures. All income statement variables use the year figures. Cash includes bank deposs and short-term financial investments. Firm size is in December 2002 US Dollars, adjusted by inflation using CPI for US firms and IGP-DI and foreign exchange rate of R$ 2.909/USD for Brazilian firms. Variations in net fixed assets and in capal investments are the value of these assets in one year minus their values in the previous year. Market equy value is calculated using the average stock quotations and the number of outstanding shares in October, November and December of each year. 5

7 The applied regression models are the Fixed Effects and the Random Effects Generalized Least Square (GLS) estimators wh or whout first order autoregressive (AR(1)) disturbance, as described by the equation (eq.1) below: lcash = β 0 β1lsize β 2opinc β 3arec β 4invent β 5apay β 6 β β 7 vcapinv β8crdebt β 9ltdebt β10ddivid 11mtb Ci vnetfa e (eq.1) where: e = ρ e u 1 in regressions wh AR(1) disturbance ( i = firm; t = year) and ρ, 1 < ρ < 1, is the autoregression coefficient of the error term, assumed to be a constant for the entire period; and, Ĉi is a constant for each firm i in the Fixed Effects model and a random variable in the Random Effects model. It is included the year dummies in all models. AR(1) disturbance is included, considering the problem of autocorrelation in residuals faced by other authors in previous researches. It is also applied the Ordinary Least Square estimator, although we are aware of s limation, for comparison purposes. The best model specification among these five models is to be chosen applying the Breusch and Pagan Lagrangian multiplier test for random effects (to select between the Ordinary Least Square and Random Effects models), the Hausman specification test (to select between the Random Effects and Fixed Effects models) and checking the relevance of the autocorrelation (to select between the model wh or whout AR(1) disturbance). B) Differences in the determinants between Brazil and the US In order to test whether the coefficients of Brazilian firms are statistically different from those of the US firms, is applied the F test, by which is tested the joint statistical significance of the coefficients on the country dummy (taking on 1 if the firm is a Brazilian and 0 if is a US firm) and s cross interactions wh the explanatory variables, using the equation (eq.2), as described below. 5 It is also included the year dummies in the model. lcash = β 0 β lsize 1 β opinc 2 β arec 3 β4invent β5apay β6vnetfa β 7vcapinv β8crdebt β9ltdebt β10ddivid β11mtb β d BRlsize 12 β d BRopinc 13 β d BRarec 14 β15d BRinvent β16d BRapay β d BRvnetfa 17 β d BRvcapinv 18 β19d BRcrdebt β 20d BRltdebt β d ddivid β d mtb d C e (eq.2) 21 BR 22 BR BR i where: and e ρ e u 1 in regressions wh AR(1) disturbance = d BR is the country dummy. C) Differences in the determinants over time The regressions using the equation (eq.3), another unrestricted version of (eq.1) wh the year dummies and their cross interactions wh the explanatory variables, as described below, wipe the effects of the different years off from the coefficients on explanatory 6

8 variables, leaving them as the coefficients of the base year (e.g., the year whout s correspondent dummy in a equation wh all other year dummies; the (eq.3) illustrates the case of 1995 as the base year). lcash = β 0 β lsize 1 β opinc 2 β arec 3 β4invent β5apay β6vnetfa β 7vcapinv β8crdebt β9ltdebt β10ddivid β11mtb β d lsize β d opinc β14d96arec β22d96mtb β lsize β d opinc β d arec β d mtb 23d β d lsize β d opinc β d arec β d mtb C e (eq.3) i where: e ρ e u 1 in regressions wh AR(1) disturbance = d95 d96 d02 and,,, are the year dummies for 1995, 1996,, And, similarly to the effect of the country, the F statistic resulted from the joint significance test of the coefficients on a specific year dummy and s cross interactions wh the explanatory variables, considering the previous year as the base year, indicates the existence or not of a structural change of that year in relation to the previous year. D) Interpretation of the regression coefficients We investigate the possible reasoning for a variable to be a determinant, verifying the applicabily of the tradeoff and the buffer stock liquidy theory. Under regression settings, we can consider the predictions for each possible explanatory variable, controlling for other variables, e.g., keeping other variables constant. A problem is, however, that there are no clear-cut predictions in terms of posive or negative relationship wh the cash holdings for many of the variables under two theories and, therefore, the underlying decision rationale may not be completely distinguishable. IV. Data The data were collected from Standard & Poor s Compustat Global Industries and Issues provided by Wharton Research Data Service, from which we obtained the raw data for 2724 US firms, and Economática (a private capal market information service provider in Brazil), from which we obtained the raw data for 396 Brazilian firms, from 1994 to It was included only the data on non-financial firms in the sample. The firm size in 2002 US dollars were calculated correcting them for inflation (CPI in the US and IGP-DI in Brazil) and, for the Brazilian firms, applying also the foreign exchange rate of R$ 2.909/US$, the average rate in The variations in net fixed assets were calculated subtracting the value of net fixed assets of one year from the previous year, and the similar proceeding was applied to the variation in capal investments; and, this procedure made us not to include the year 1994 in the regressions. To calculate the market-to-book ratio, we utilized the market stock price and the outstanding number of shares, considering the average of the last three months of the year, when available. In order to avoid the influence of outliers, the observations below 1% percentile in opinc, vnetfa and vcapinv and above 99% percentile in cash, opinc, arec, invent, apay, vnetfa, vcapinv, crdebt, ltdebt and mtb were dropped from the 7

9 sample. It was also dropped firms wh only one year of observation. And, was randomly selected 250 firms for each country in order not to have the influence of the number of the cross sections in the comparison between two countries. The basic descriptive statistics of the resulted sample are presented in the table 2. The values in the table show that, in average, in the period of study: The US firms had greater cash holdings than Brazilian firms in the analyzed period: 42.5% of total non-cash assets in the US and 8.7% in Brazil. 6 The average firm size of 3,666 million dollars in the US was also much larger than the average of 455 million dollars in Brazil. A surprising figure is the average operating income: was negative in the US, while was posive in Brazil. Proportionally to the non-cash total assets, the US firms invested more in the net fixed assets than the Brazilian firms; however, the Brazilian firms made more capal investments than the US firms. The account receivables, inventories and account payables of US firms were all greater than those of the Brazilian firms, indicating that there were some differences in working capal management between two countries. The US firms used mainly the long-term debt for debt financing: represented 24.7% of total non-cash assets, whereas the current debt represented 4.5%; on the other hand, the Brazilian firms relied more on the current debt, although s average figure was not so greater than the long-term debt for the firms in the sample: the current debt represented 12.6% of non-cash assets versus 12.1% of long-term debt. Comparing the total debt (the current plus long term-debt), can be noted also that the US firms were slightly more leveraged than the Brazilian firms: 29.2% for the US firms and 24.7% for the Brazilian firms; however, one must take into consideration in this difference that the financial leasing is included as debt on the balance sheets in the US, but not in Brazil. A larger proportion of firms paid out dividends in the US in comparison to Brazil. The average market-to-book ratio was also much greater in the US than in Brazil. Table 2. Descriptive Statistics of the Sample US (# of firms = 250) BRA (# of firms = 250) Variable Mean Std. Dev. Min Max Mean Std. Dev. Min Max cash size opinc vnetfa vcapinv arec invent apay crdebt ltdebt ddivid mtb Note: See the Table 1 for the variable definions. 8

10 V. Results A) The determinants of corporate cash holdings in Brazil and the US The results of the regressions are presented in the table 3. The F statistics for OLS, FE and FE wh AR(1) models and χ 2 statistics for RE and RE wh AR(1) models indicate that all of them have statistically significant explanatory powers. Table 3. Results of Regressions for the period, applying Ordinary Least Squares, Fixed Effects, Fixed Effects wh AR(1) Disturbance, Random Effects, and Random Effects wh AR(1) Disturbance Dependent Variable = lcash Explanat. US BRAZIL variable OLS (a) FE FE AR(1) RE RE AR(1) OLS (a) FE FE (AR1) RE RE (AR1) lsize (5.68)** (0.06) (3.18)** (3.49)** (3.61)** (4.66)** (0.54) (1.31) (3.84)** (4.02)** opinc (11.77)** (7.74)** (5.10)** (13.75)** (11.76)** (5.27)** (2.95)** (2.75)** (4.01)** (4.33)** vnetfa (2.18)* (1.67) (1.52) (3.09)** (1.15) (3.71)** (4.03)** (3.48)** (4.58)** (4.85)** vcapinv (0.58) (0.89) (1.17) (0.16) (0.84) (0.49) (1.03) (1.01) (0.84) (0.78) arec (2.46)* (3.96)** (4.76)** (3.56)** (3.90)** (0.75) (1.63) (0.08) (0.18) (0.36) invent (4.39)** (1.05) (0.58) (3.16)** (2.91)** (7.86)** (7.50)** (4.42)** (8.96)** (8.03)** apay (4.46)** (3.04)** (3.64)** (4.05)** (4.33)** (0.64) (0.42) (2.00)* (0.38) (0.17) crdebt (6.60)** (2.72)** (1.32) (4.54)** (3.26)** (1.91) (5.55)** (5.39)** (3.73)** (4.01)** ltdebt (0.03) (1.44) (3.36)** (1.72) (2.63)** (1.52) (4.06)** (2.83)** (2.88)** (2.74)** ddivid (5.95)** (0.14) (0.78) (2.95)** (2.83)** (2.28)* (1.58) (0.12) (0.19) (0.33) mtb (7.23)** (6.52)** (3.05)** (8.54)** (8.94)** (0.35) (0.82) (1.27) (0.81) (0.68) constant (7.18)** (1.25) (4.33)** (5.96)** (5.67)** (2.29)* (0.49) (1.53) (2.01)* (2.28)* ρ n.a. n.a n.a n.a. n.a n.a DW Statist n.a. n.a n.a n.a. n.a n.a F or χ 2 Stat P-value # of observ # of firms Notes: See the Table 1 for the variable definions. Absolute value of z statistics for Random Effects regressions and t statistics for Ordinary Least Squares and Fixed Effects regressions in parentheses: * significant at 5%; ** significant at 1%. (a) Whe heteroskedasticy-consistent t statistics. ρ is the autoregression coefficient of the AR(1) disturbance term in the regression equation. F Statistics for OLS, FE and FE (AR1) regressions and χ 2 Statistics for RE and RE AR(1)s regressions, wh respective P-values. DW Statist is modified Bhargava et al. Durbin-Watson statistics. 9

11 The Breusch and Pagan Lagrangian multiplier tests for random effects result in the χ 2 statistics of (P-value of 0.000) for the US firms and (P-value of 0.000) for the Brazilian firms, rejecting s null hypothesis that the variance of the Ĉ i term in the equation (eq.1) is equal to zero and, hence, that the Random Effects models are more adequate than the pooling regressions by the OLS model for both countries. The Hausman specification tests result in χ 2 statistics of (P-value of 0.000) for the US firms and 80.4 (P-value of 0.000) for the Brazilian firms, rejecting s null hypothesis that the differences in coefficients are not systematic and not condional on the Ĉ i term, indicating that the Fixed Effects models are more appropriate than the Random Effects models. In fact, the hypothesis that the term C i is not correlated wh the explanatory variables, which is a key supposion of the Random Effects models to result in unbiased and consistent estimates, seems to be too strong in our case. Moreover, the existence of autocorrelation in residuals is clear wh the modified Bhargava et al. Durbin-Watson statistics of for the US firms and for the Brazilian firms, as well as the autoregression coefficient ρ of for the US firms and for the Brazilian firms. Thus, we conclude that the Fixed Effects model wh AR(1) disturbance is the most suable among the five models for our analyses. According to the results of the selected model, in the US, the firm size, account receivables, account payables, long-term debt and market-to-book ratio are the statistically significant determinants wh posive effects on the cash holdings, whereas the operating income is the statistically significant determinant wh negative effect. Some of these results are different from those reported by OPSW (1999) for the period, according to whom the firm size and debts have negative effects on the cash holdings in the US. In Brazil, on the other hand, the results indicate that the operating income, inventories, account payables, current debt and long-term debt are the determinants wh posive effects on the cash holdings, whereas the variation in net fixed assets is the determinant wh negative effect, which are all consistent wh the results reported by KC (2003). B) Differences in the determinants between Brazil and the US We can verify the differences in the coefficients between Brazil and the US, comparing them in the table 3. In fact, verifying wh scientific rigor, the F statistic of the coefficients on the country dummy and s cross interaction wh explanatory variables of 3.47 (P-value of 0.000), resulted from the regression applying the Fixed Effects model wh AR(1) disturbance, using the equation (eq.2), confirm that the differences in the coefficients between the US and Brazil are statistically significant. C) Differences in the determinants over time The results of the regressions in each country, considering each year from 1995 to 2002 as the base year (e.g., the year whout the correspondent dummy), applying the Fixed Effects model wh AR(1) disturbance, and using the equation (eq.3), are presented in the table 4. These results are different from the results of the same model reported in the table 3, due to the inclusion of the cross interactions of the year dummies wh the explanatory variables. 10

12 Table 4. Results of Regressions Year by Year from 1995 to 2002, applying Fixed Effects wh AR(1) Disturbance Model and Cross Effects between Year Dummies and all Independent Variables Dependent Variable = lcash Explanat. Predicted Sign US BRAZIL variable Tradeoff Buff. St lsize (3.81)** (4.68)** (5.37)** (5.16)** (5.13)** (5.39)** (5.42)** (5.18)** (3.51)** (3.78)** (3.90)** (3.18)** (2.06)* (1.66) (1.11) (0.63) opinc (4.48)** (2.39)* (2.00)* (3.17)** (3.72)** (1.22) (2.38)* (1.70) (0.99) (2.35)* (1.85) (1.35) (0.51) (0.45) (0.54) (3.28)** vnetfa (0.15) (1.38) (0.66) (2.00)* (0.31) (3.18)** (3.39)** (2.10)* (1.58) (2.38)* (1.30) (4.74)** (1.08) (0.53) (1.24) (4.03)** vcapinv (0.44) (0.37) (1.75) (0.64) (1.53) (3.83)** (1.76) (1.00) (1.44) (1.85) (2.23)* (2.32)* (0.91) (0.04) (1.37) (0.65) arec or (0.37) (0.34) (1.89) (3.53)** (4.21)** (5.81)** (6.70)** (5.00)** (0.52) (1.22) (2.52)* (0.78) (0.42) (0.32) (0.09) (0.60) invent or (2.43)* (2.21)* (1.68) (1.10) (1.60) (1.42) (0.71) (1.84) (0.73) (0.24) (3.40)** (3.58)** (1.89) (2.12)* (3.34)** (1.43) apay or (0.12) (0.12) (1.05) (1.50) (2.59)** (2.41)* (2.42)* (3.04)** (0.59) (0.34) (0.32) (1.18) (2.09)* (2.35)* (2.40)* (1.82) crdebt or (1.18) (1.68) (1.89) (0.64) (1.06) (0.19) (0.39) (2.64)** (2.64)** (0.59) (1.05) (2.77)** (3.16)** (3.95)** (2.97)** (2.06)* ltdebt or (0.27) (0.66) (1.68) (0.08) (2.83)** (7.67)** (3.48)** (1.66) (0.50) (0.11) (0.25) (3.46)** (1.87) (3.12)** (2.84)** (1.89) ddivid or (0.14) (0.53) (1.16) (1.36) (0.03) (1.06) (1.29) (1.25) (1.04) (0.02) (0.08) (0.96) (0.21) (0.88) (0.36) (0.42) mtb or none (5.65)** (3.38)** (0.54) (0.24) (2.83)** (4.10)** (0.44) (0.31) (0.38) (0.22) (1.99)* (0.51) (0.24) (1.69) (1.75) (0.37) constant (3.60)** (3.60)** (3.60)** (3.60)** (3.60)** (3.60)** (3.60)** (3.60)** (1.21) (1.21) (1.21) (1.21) (1.21) (1.21) (1.21) (1.21) ρ DW Statist F Statist P-value # of observ # of firms Notes: See the Table 1 for the variable definions. Absolute value of t statistics in parentheses: * significant at 5%; ** significant at 1%. (a) Whe heteroskedasticy-consistent t statistics. ρ is the autoregression coefficient of the AR(1) disturbance term in the regression equation. F Statistic and P-value for joint significance of all coefficients incluing all year dummies. DW Statist is modified Bhargava et al. Durbin-Watson statictics. 11

13 Again, we can observe the differences in the coefficients over time both in the US and in Brazil. And, the F statistics of the coefficients on the year dummy and s cross interaction wh the explanatory variables, presented in the table 5, confirm statistically that the coefficients are jointly different from those of the previous year in all years. These differences in the coefficients indicate that there are constant structural changes over time in the determinants of corporate cash holdings in two countries, except in 2002 in the US and except in 2000 and 2001 in Brazil, when is not possible to reject the hypothesis that these differences are equal to zero. The non-statistical significance of the changes in the coefficients in Brazil in these two years may be related to some dominant influencing factors associated wh the determinants of corporate cash holdings, probably related to the economic suation after the Real devaluation in D) Interpretation of the regression coefficients The interpretation of the determinants, analyzing them year by year, gives us many insights about the differences and similaries in the management of cash holdings in two countries. To begin wh, the posive effect of the firm size is the same in Brazil and in the US, whout change in the sign from one year to another, which is consistent wh the buffer stock liquidy theory, although s economic significance changes and is not significant in Brazil since The sign on the operating income in the US is posive in 1995 and 1996, consistently wh the buffer stock liquidy theory, and turns to negative since 1997, although not statistically significant in 2000 and 2002, consistently wh the tradeoff theory; in Brazil, on the other hand, is posive, wh statistical significance in 1996 and 2002, consistently wh the buffer stock liquidy theory. In the US, the sign on the variation in net fixed assets is negative and statistically significant in 1998 and from 2000 to 2002, and the sign on the variation in capal investments is posive and significant in In Brazil, the sign on the variation in net fixed assets is generally negative (except in 1995, when is posive but not statistically significant) wh statistically significant coefficients in 1996, 1998 and 2002 and the sign on the variation in capal investments is negative and statistically significant in 1997 and posive and statistically significant in These results indicate that, in terms of the variations in net fixed assets and capal investments, firms behaviors follow the buffer stock theory most of the time; however, they follow the speculative motive of the tradeoff theory in some specific years (2000 in the US and 1998 in Brazil). Table 5. F Statistics to Analyze the Changes in the Determinants Over Time US (0.00) (0.00) (0.01) (0.01) (0.00) (0.00) (0.16) BRAZIL (0.00) (0.00) (0.00) (0.00) (0.29) (0.04) (0.00) Notes: F statistics of coefficients on the year dummy and s cross interactions wh the explanatory variables are obtained using the regression equation wh all year dummies except that of the previous year. P-values in parentheses. 12

14 In terms of working capal variables in the US, the inventories are the economically and statistically significant variables in 1995 and 1996, but the accounts receivables and payables are more significant from 1998 to In Brazil, the more significant variables are inventories from 1997 to 2002 and also the account payables from 1999 to The posive sign on the inventories indicates that the firms follow the precaution motive under the tradeoff theory. For instance, is consistent wh the suation of firms facing troubles in sales growth, in spe of holding inventories prepared for the growth. The posive sign on the account payables is consistent wh the suation of harder access to the financial markets and firms seeking alternative financings from the suppliers, under the buffer stock liquidy theory; however, is also consistent wh the precaution motive related to the suppliers under the tradeoff theory. The sign of the coefficient on the account receivables is generally posive in the US (except in 1995, when is negative but not statistically significant) and statistically significant since 1998, which indicates the tradeoff based decision wh precaution motive related to bad receivables. In Brazil, the account receivables are not statistically significant in general; however, s negative and statistically significant coefficient in 1997 indicates probable higher importance given to the sales on cred particularly in that year, meeting the demand for working capal financing from the clients, consistently wh the buffer stock theory; however, s reasoning is indistinguishable wh the substution effect under the tradeoff theory. The current debt is not statistically significant in the US, except in 2002, when is negative and statistically significant, and the long-term debt is posive and significant from 1999 to In Brazil, the sign on current debt is negative and statistically significant in 1995, and both current and long-term debts are posive and statistically significant from 1998 to 2002 (although the long-term debt is not statistically significant in 1999 and 2002). These years were when the instabily in the financial markets increased due to the turmoil in the emerging markets; and, the posive coefficient sign (meaning that firms wh higher debt tend to hold higher cash) can be explained by the need of higher cash buffers for the transaction and precaution motives under the tradeoff theory, former applying more for the long-term debt and the latter for the current debt. In addion, is very likely that Brazilian firms held addional cash for the arbrage motive in some of the years. Nevertheless, the actual reason is indistinguishable wh the buffer stock liquidy theory since the predicted sign is the same. The effect of the dividends is not significant both in the US and in Brazil, which is non-consistent neher wh the tradeoff theory nor wh the buffer stock liquidy theory. The effect of the market-to-book ratio is also not significant in Brazil, which is consistent wh the buffer stock liquidy theory, except in 1997, when is posive and significant. In the US, s effect is negative and significant in 1995 and 1996 and is posive and significant in 1999 and Its negative sign is consistent wh the tradeoff theory wh the market-to-book ratio as a proxy for the access to financial markets; and, s posive sign is consistent wh the tradeoff theory wh the market-to-book ratio as a proxy for the growth opportuny. Overall, our results indicate that there are similaries between Brazil and the US in terms of the effects of firm size, investments, inventories, account payables and long-term debt on the cash holdings; however, there are differences between two countries in terms of the effects of operating income, account receivables, current debt and market-to-equy ratio. It cannot be totally certain in all cases; however, is very likely that the tradeoffs involve 13

15 different factors and rationale in Brazil and in the US. Also, there seems to be some relation between the determinants of cash holdings and the changes in the economic and financial environment. In terms of the applicabily of the tradeoff and the buffer stock theories, seems that neher the one nor the other individually explains the determinants of corporate cash holdings appropriately. On the contrary, is more reasonable to treat them as complementary instead of opposing theories. Under this combined view, is applied the tradeoff theory for some variables and the buffer stock liquidy theory for others, depending on their importance in a particular country in a particular moment in time, to explain the firms behaviors in relation to the cash holdings. VI. Conclusion In this paper we compared the determinants of cash holdings in Brazil and in the US for the period. Our results indicate that some of the determinants and the tradeoffs associated wh these determinants in Brazil are different from those in the US. Also, many of the determinants and their tradeoffs change over time both in Brazil and the US, probably following some macroeconomic factors. In addion, we conclude that the determinants of corporate cash holdings can be explained partially but not totally by the tradeoff theory in two countries and that the decision taking process should be regarded as a combination of active cash holding management, based on the tradeoff between benefs and costs, wh the passive management, letting the cash holdings be the buffer stock of cash inflows from sources of funds and outflows to uses of funds, at the same time. We should be aware, however, of the limation of the results. One limation is the biases caused by the endogeney among the analyzed variables, which are basically accounting data, an internally closed system, in our case. The dynamic panel data regression model of Arellano and Bond (1991), applied by Ozkan and Ozkan (2004) and Guney, Ozkan and Ozkan (2003), which uses lagged first differences as instrumental variables, is not a good solution in our case, since the lagged first differences may not be correlated wh the original explanatory variables and turn out not to be adequate instruments when there are structural changes over time. Another limation is the supposion of unidirectional effect of explanatory variables on the cash holdings, which may not conform to the realy. Indeed, the cash holdings are very likely to be affected by working capal, investment, financing, dividend and market variables; however, the effects in the oppose direction are also very likely to exist. The differences in accounting methods between the US and Brazil, such as the inclusion of financial leasing as debt in the former country and not in the latter, making the comparison between firms in these two countries not straightforward, are another limation. Regression models which migate the endogeney problem and consider the bidirectional effects, such as system of simultaneous equations, may give us clearer conclusions. Further research opportunies include the analyses of the following effects on the cash holdings: ownership structure, foreign sales and purchases, and loans in foreign currency. The effect of corporate cash holdings on the firm performance and value is also an interesting future research subject. 14

16 References Arellano, M. and Bond, S.; Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies 58, Anderson, R.W.; Capal structure, firm liquidy and growth. Working paper no.27, National Bank of Belgium. Bruinshoofd, W.A. and Kool, C.; The determinants of corporate liquidy in the Netherlands. Working paper, Maastricht Universy. Dtmar, A.; Smh, J.M. and Servaes, H.; International corporate governance and corporate cash holdings. Journal of Financial and Quantative Analysis 38:1, Falls, G.A. and Natke, P.A.; Cash flow instabily and the demand for liquid assets by firms in Brazilian manufacturing. Quarterly Review of Economics and Finance 36-2, Guney, Y.; Ozkan, A. and Ozkan N.; Addional international evidence on corporate cash holdings. Working paper, Universy of York. Harford, J.; Corporate cash reserves and acquisions. Journal of Finance 54, No.6, Kim, C.S.; Mauer, D.C. and Sherman, A.E.; The determinants of corporate liquidy: theory and evidence. Journal of Financial and Quantative Analysis 33-3, Keynes, J.M.; The general theory of employment, interest and money. London: Harcourt Brace. Koshio, S. and Cia, J.N.; The determinants of corporate cash holdings in Brazil. Unpublished working paper, São Paulo, Fundação Getúlio Vargas-EAESP. Myers, S.C.; The capal structure puzzle. Journal of Finance 39-3, Myers, S.C. and Majluf, N.; Corporate financing decisions when firms have information that investors do not have. Journal of Financial Economics 13, Natke, P.A.; The firm demand for liquid assets in an inflationary environment. Applied Economics 33, Opler, T.; Pinkowz, L.; Stultz, R. and Williamson, R.; The determinants and implications of corporate cash holdings. Journal of Financial Economics 52, Ozkan, A. and Ozkan, N.; Corporate cash holdings: an empirical investigation of UK companies. Journal of Banking and Finance, forthcoming. Pinkowz, L.; The market for corporate control and corporate cash holdings. Working paper, Georgetown Universy. 15