Journal of Banking & Finance

Journal of Banking & Finance 36 (2012) 26 35 Contents lists available at ScienceDirect Journal of Banking & Finance journal homepage: www.elsevier.com/locate/jbf Cash holdings in private firms Marco Bigelli a,, Javier Sánchez-Vidal b a Department of Management, University of Bologna, Via Capo di Lucca 34, Bologna 40126, Italy b Department of Finance and Accounting, Technical University of Cartagena, Calle Real, 3, Cartagena 30201, Spain article info abstract Article history: Received 7 December 2009 Accepted 7 June 2011 Available online 23 June 2011 JEL classification: G31 G32 Keywords: Cash holdings Cash determinants Private firms Trade-off model Pecking order theory Evidence from a wide sample of Italian private firms shows that cash holdings are significantly related with smaller size, higher risk and lower effective tax rates, therefore supporting predictions from the trade-off model. More cash is also held by firms with longer cash conversion cycles and lower financing deficits, as predicted by the financing hierarchy theory. Reported evidence also shows that dividend payments are associated with more cash holdings, and both bank debt and net working capital represent good cash-substitutes. When controlling for macroeconomic and industry factors, some variables lose their significance, but the general findings are confirmed. Finally, cash-rich companies are found to be more profitable, to pay more dividends and to invest more in a medium-term future horizon. Ó 2011 Elsevier B.V. All rights reserved. 1. Introduction Academic literature on cash holdings can be dated back to Keynes (1936), who indicates two main benefits from holding cash: lower transaction costs from not having to liquidate assets when facing a payment and a valuable buffer to meet unexpected contingencies. The literature about cash specifically applied to companies is generally traced back to Miller and Orr (1966), who develop a trade-off model for the determination of the optimal level of cash holdings by balancing the costs of running out of cash and the costs of holding non-interest bearing cash. The trade-off model of optimal cash holdings (Miller and Orr, 1966; Kim et al., 1998) is typically opposed to the financing hierarchy theory (Myers and Majluf, 1984), which does not assume an optimal level and expects higher levels of cash reserves in more profitable firms as a financial slack. Together with these two main views, there are several other hypothesis that contribute to the determinants of cash holdings. In fact, financially constrained firms, i.e. firms with a lower access to external financing, should have a higher propensity to save cash out of cash flows (Almeida et al., 2004), should prefer cash to lower debt for higher levels of hedging needs (Acharya et al., 2007) and have a higher dollar value of cash held (Faulkender Corresponding author. Tel.: +39 051 2098060; fax: +39 051 6390612. E-mail addresses: marco.bigelli@unibo.it (M. Bigelli), javier.sanchez@upct.es (J. Sánchez-Vidal). and Wang, 2006). When agency costs from ownership control separation are relevant (as in the US), cash-rich firms are more likely to engage in value destroying acquisitions (Harford, 1999), shareholders assign a lower value to cash holdings in diversified firms (Tong, 2009), in firms with a higher degree of information asymmetry (Drobetz et al., 2010) and in firms with poor corporate governance (Dittmar and Mahrt-Smith, 2007), while firms with weaker corporate governance hold smaller cash holdings (Harford et al., 2008). Since cash, as well as being wasted, can also be diverted to controlling shareholders private benefits, in countries with poor investor protection firms tend to hold more cash (Dittmar et al., 2003), though the value of one dollar of cash is lower (Pinkowitz et al., 2006) and a firm s value is lower when controlling managers hold more cash (Kalkeva and Lins, 2007). Finally, taxes on cash remittances exert a positive influence on the level of cash holdings abroad by US multinational corporations (Foley et al., 2007). The empirical analysis of the determinants of firms cash holdings has received growing attention by academics only in the last 10 years. However, existing empirical literature mainly refers to US listed companies (Kim et al., 1998; Opler et al., 1999; Harford et al., 2008; Mello et al., 2008), to the evolution of their cash holdings (Bates et al., 2009), to US multinational firms (Foley et al., 2007) and to cross-country comparisons (Guney et al., 2007; Dittmar et al., 2003; Pinkowitz and Williamson, 2001; Pinkowitz et al., 2006). Some empirical evidence has also been reported for EMU listed firms (Ferreira and Vilela, 2004) or EMU large firms 0378-4266/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jbankfin.2011.06.004

M. Bigelli, J. Sánchez-Vidal / Journal of Banking & Finance 36 (2012) 26 35 27 (Pal and Ferrando, 2010). European single country evidence is limited to the influence of managerial ownership on cash holdings of UK listed firms (Ozkan and Ozkan, 2004), the adjustment of large Dutch firms to long-run liquidity targets (Bruinshoofd and Kool, 2004), the role of intra-group relations in the cash reserves of large Belgian firms (Deloof, 2001) and the determinants of cash levels for Spanish SME firms (García-Teruel and Martínez-Solano, 2008). To the best of our knowledge, little attention has been given to the determinants of cash holdings in private firms, which should be characterized by lower or no agency costs of free cash flows (Jensen, 1986). The present paper contributes to the literature on cash holdings by analyzing the determinants of cash reserves in a wide sample of Italian private firms which are very representative of the entire population in the 1996 2005 period (17,165 firms for 152,141 firm-year observations). Results show that cash holdings by Italian firms average about 10% of total assets, comparable with the 8 10.5% level found in US listed companies (Kim et al., 1998; Opler et al., 1999; Foley et al., 2007; Mello et al., 2008) and the 12.6% level found by Drobetz et al. (2010) as an overage for 45 countries in almost the same period (1995 2005). Smaller firms, which are younger, riskier and presumably more financially constrained, hold significantly more cash and less cash substitutes than the biggest firms. Reported evidence from a panel data analysis supports both the trade-off model and the financing hierarchy theory. With respect to the former one, we find that smaller firms with riskier cash flows and lower effective tax rates hold more cash. In addition, shorter cash-conversion cycles and higher financing deficits lead to lower cash balances, thus supporting the financing hierarchy theory. Private firms that pay dividends hold more cash. Evidence also shows that both bank debt and net working capital can be considered as cash substitutes. When macroeconomic and industry factors are taken into account some variables lose part of their influence but this does not alter the general conclusions. Further evidence shows that cash rich companies tend to be more profitable, pay more dividends and invest more in a medium-term future horizon. The rest of the paper continues as follows. Section 2 discusses and adapts previous theories to a private firm context. Section 3 describes the dataset, the variables and the methodology. Section 4 analyzes the results of the statistical analysis, while Section 5 reports the main conclusions. 2. Theory and hypothesis In this section we review the main theoretical contributions on the determinants of firms cash holdings trying to check whether they hold or whether they should be adapted in a private firm context. Due to the wider amount of available information, most academic literature concentrates on public firms though private firms internationally outnumber public firms and represent a much greater proportion of total assets. According to the Bank of Italy, in 2001 there were more than 4 million Italian private firms (of which 116,000 were limited companies) employing more than 97% of the total workforce, versus only 287 listed firms. Even where financial markets are more developed, as in the UK, private firms represent 97.5% of total incorporated entities and more than two-thirds of total assets (Brav, 2009). One fundamental difference between private and public firms is the ownership structure. Public firms have thousands of shareholders while private firms have just one or few shareholders, often belonging to one single family. In a sample of Italian middle and large private firms, about two-thirds of the firms have between one and four shareholders while the median number is three (Giacomelli and Trento, 2005). In the UK, one-third of private firms have only one shareholder and the median is two (Brav, 2009). Using data from the World Bank s Enterprise Survey, Hope et al. (2010) report that in the 2002 2005 period, in a sample of about 31,000 private firms over 68 countries, the largest shareholder owns on average a 74% stake. A second relevant difference is that private firms are more reluctant to access equity capital markets because of a desire to retain control over the firm as well as the higher cost of equity due to their greater level of opacity (Brav, 2009). As a consequence, private firms are typically characterized by higher leverage ratios and a greater proportion of short-term debt to total debt. Consistent with their less frequent access to capital markets, Brav (2009) also finds that private firms stockpile cash in good times and dilute cash reserves in bad times. Moreover, compared to public firms, private firms tend to increase their investments with a lag from a profitability increase. 2.1. The trade-off model of cash holdings In a world of perfect capital markets there would be no transaction costs for raising cash, and holdings of liquid assets would not affect a firm s value. But markets are far from perfect and transaction costs are never irrelevant. Companies must therefore determine the optimal level of cash holdings by trading off the marginal cost of holding liquid assets (lower return) with its marginal benefit (minimization of transaction costs, undertaking of investment opportunities in case of market frictions, etc.) as suggested by Miller and Orr (1966) and Kim et al. (1998). The tradeoff model is typically tested using the variables defined below. 2.1.1. Size Since there are substantial fixed costs of acquiring outside financing as well as economies of scale in cash management (Dittmar et al., 2003) both in public and private firms, larger companies are expected to get financing in an easier and cheaper way. In addition, raising cash by selling non-core assets in periods of financial distress (Lang et al., 1995) should be easier for diversified companies, and large companies tend to be more diversified (Rajan and Zingales, 1995). From the above relationships, the trade-off model would predict lower cash holdings in larger private firms. 2.1.2. Risk or volatility of cash flows Cash can be considered a buffer to absorb adverse shocks and increase the probability of survivorship during periods of poor business conditions. The precautionary motive for cash holdings is also related to potential concerns about having to cut dividends or suffer potential losses from forced divestitures of assets to obtain cash. It is therefore common sense to think that higher levels of uncertainty and risk are typically associated with higher levels of cash reserves, especially for financially constrained firms (Han and Qiu, 2007). In private firms, the level of financial frictions should be higher and the access to external financing is more difficult. In addition, the cost of failure is also higher because of legal, accounting, trustee and auctioning fees (Ang, 1991). We therefore expect such a relationship to be even stronger. Hence, more variable cash flows should lead to higher levels of cash holdings in private firms. 2.1.3. Effective tax rate If a company exhibits a high cash figure, which could otherwise be used for reducing debt or buying back shares, the opportunity cost of cash holdings will increase as the firm s tax rate increases (Opler et al., 1999). In fact, higher tax rates would correspond to higher tax shields from debt and higher opportunity costs for holding more cash and less debt. We therefore expect that private firms with higher effective tax rates hold lower cash balances.

28 M. Bigelli, J. Sánchez-Vidal / Journal of Banking & Finance 36 (2012) 26 35 2.1.4. Growth opportunities If a company has future valuable investment opportunities, it will try not to run out of cash by the time it needs it. Besides, the literature on real options has emphasized the option value of waiting to invest (McDonald and Siegel, 1986). For these reasons, holdings of liquid assets could be more valuable for firms with more investment opportunities and for firms whose set of investment opportunities changes over time, especially if they are financially constrained (Denis and Sibilkov, 2010). Such a relationship gets reinforced for private companies as they should be characterized by a greater risk of underinvestment due to a shortage of internally generated funds (Ang, 1991). We therefore expect that higher growth opportunities are associated with higher cash holdings in private firms. 2.2. The financing hierarchy theory and cash holdings In the presence of information asymmetries, according to the pecking order theory (Myers and Majluf, 1984), companies prefer to finance their investments with retained earnings, then with debt and, lastly, with equity. This financing hierarchy should be even stronger in Italian private firms, both because Italian private and public equity markets are relatively underdeveloped (Pinkowitz et al., 2006) and because information asymmetries in private firms are certainly higher than in listed firms, being subject to lower levels of disclosure, supervision and external auditing. The financial hierarchy theory is typically associated with the following variables. 2.2.1. Financing deficit According to the pecking order theory, there is neither an optimal debt for the company nor an optimal level of cash holdings. It follows that when internal funds are not enough, companies will raise debt and leverage. The firm s cash flow is therefore typically adopted as an explanatory variable, though with different definitions (Opler et al., 1999; Ozkan and Ozkan, 2004). We think that the firm s financing deficit could be a more comprehensive variable as it not only includes the firm s cash flow but it measures the exact amount of the firm s yearly external financing (Shyam-Sunder and Myers, 1999). A firm s financing deficit would then lead either to a reduction of cash holdings or to an increase in debt, or both. We therefore expect that private firms with higher financing deficits have lower cash holdings and higher debt levels. 2.2.2. Dividends payment When private firms need to obtain cash and have difficulties in raising more debt, either because they are financially constrained or because they are already too leveraged, they may try to increase retained earnings by cutting dividend payments. Financially constrained private firms should therefore try to obtain their cash needs by cutting up dividends (Fazzari et al., 1988). In most of the private firms there is only one shareholder, often the founder or the founding family (Corbetta and Montemerlo, 1999; Brav, 2009). A firm s assets and personal wealth are often perceived as a whole, especially in small firms (Ang, 1992). It follows that the choice to pay out dividends depends mostly on the firm s needs for retaining cash to finance new investments or to face a financing deficit, when it is not due to exploit some tax advantages. US public companies, instead, commit themselves to regular dividend payments in order to reduce managerial agency costs, as they will more frequently have to return to capital markets for raising new capital, with a consequent more accurate and efficient monitoring (Easterbrook, 1984). Such commitment does not characterize private companies as in most cases there is no ownership control separation and managers tend to coincide with owners. Aligned with the above thesis, Brav (2009) finds that public firms do not change their dividends much in response to changes in their performance while dividend payments in private firms are much more sensitive to their operating performance. It follows that while the typical relationship for US corporations is that firms that pay dividends tend to hold less cash balances (Opler et al., 1999; Harford et al., 2008), in private firms we could expect an opposite one, as payment of dividends in private firms should be correlated with an excess generation of cash flows, and lower or no dividends associated with a cash shortage. We therefore expect that private firms that pay dividends have higher cash balances. 2.2.3. Cash conversion cycle On the same line of reasoning, the length of the cash conversion cycle will determine the firm s ability to generate cash from ongoing operations: the shorter the length the higher the frequency of cash generation. As Opler et al. (1999) suggest, if a company manages to have a shorter cash conversion cycle it will have a lesser need of cash balances and this should hold both for public and private firms. We therefore expect that private firms with shorter cash conversion cycles exhibit lower cash balances. As anticipated above, agency arguments may help explain cash holdings of listed firms but cannot be applied to private firms. In fact, managers of listed companies may prefer to hold more cash either because they are risk-averse or because cash helps the consumption of perquisites or the financing of excess investments, as the free-cash-flow theory suggests (Jensen, 1986). However, in order to have agency costs of equity as a cash determinant, ownership and control should be separated (Jensen and Meckling, 1976), as it is often the case for US listed companies and is not likely to be the case for private companies, where ownership is fully concentrated in one single shareholder or a family or a private equity investor. Agency motives will therefore not be included in our set of cash determinants. 2.3. Financial constraints and substitutes of cash holdings Since cash balances can be used to redeem debt, cash can be generally considered as the other side of the same coin (Opler et al., 1999). As commented before, equity agency costs should be negligible in private firms, but debt agency costs could be more relevant than in public firms, as monitoring-like solutions are relatively more expensive for small businesses (Ang, 1991). Information asymmetries are particularly severe for private firms, as they lack a public price as a mechanism to aggregate investor information, they do not benefit from analyst research and they supply limited disclosures to investors (Mantecon, 2008). Higher debt agency costs and information asymmetries should lead to tighter financial constraints and make the following variables more determinant for private firms. 2.3.1. Cash as negative debt According to Acharya et al. (2007), cash can be viewed as Negative debt only in the absence of financial frictions or in constrained firms with low hedging needs. In fact, only financially constrained firms with low hedging needs (low correlation between cash flows and investment opportunities) tend to use excess cash flows to reduce debt and show a weak cash flow sensitivity of cash, i.e. a low propensity to save cash out of the firm s cash flows. However, only the data on US public firms supports these arguments and some of the variables used to measure financial constraints (bond rating or commercial paper rating) would not be applicable to private firms. Besides, the level of financial constraints that characterize private firms is not easy to determine. In fact, on the one hand, private firms should have more severe financial constraints than listed firms, being smaller in size and

M. Bigelli, J. Sánchez-Vidal / Journal of Banking & Finance 36 (2012) 26 35 29 having no direct access to capital markets. On the other hand, private firms typically have just one or few shareholders belonging to the founding family (Brav, 2009) and can easily get bank-debt if pledged by the family. These ownership ties may substitute the intra-group relations which lessen liquidity needs for Belgian listed firms in business groups (Deloof, 2001). Bank debt can actually play a major role in determining the amount of cash holdings in private firms. In fact, in order to monitor small private borrowers, banks tend to gather non-public information, and this could reduce the high information asymmetry that characterizes them (Fama, 1985). Secondly, the firms credit worthiness could be enhanced by established relationships with banks (James, 1987; Mikkelson and Partch, 1986) and this could further reduce financial constraints. More financial flexibility could also derive from the renegotiation of bank loans upon maturity (Chemmanur and Fulghieri, 1994). It follows that debt provided by bank loans should lessen financial constraints in private firms and serve as a cash substitute. The general wisdom that cash can be considered as negative debt should therefore hold and private firms with higher levels of bank debt are expected to have lower cash balances. 2.3.2. Net working capital as a cash substitute The other typical substitutes of cash are bank lines of credit (Demiroglu and James, 2011) or those current assets that can easily be transformed into cash. Receivables, for example, can be easily cashed out through factoring in small firms or by securitization processes in larger ones. Net working capital (net of cash) can therefore be considered as a cash substitute (Opler et al., 1999) both in public and private firms. It follows that private firms with a higher net working capital should have lower cash balances. 3. Sample, variables definition and methodology From the Bureau Van Dijk AIDA database we initially take all Italian unlisted companies which have detailed financial statements (as we are looking for disaggregated data for both cash and cash equivalents) in every year in the 1996 2005 period. The initial database is made up of 19,163 firms and 191,630 firm-year observations. We eliminate those firms whose 2005 turnover was below 2 million euros (considered micro firms by EU definition). As Opler et al. (1999), we exclude financial firms and eliminate utility firms as their cash holdings can be subject to some form of regulatory supervision. We then exclude those firm-year observations reporting either negative sales or a ratio of cash on total assets greater than one or lower than zero. Similar to Almeida et al. (2004) and Acharya et al. (2007), we exclude those firm-year observations reporting a yearly change in total assets greater than 100%. 1 Finally, we exclude those firm-year observations reporting a negative equity of the firm. The final sample is composed by 17,165 firms having an average of 8.9 years per company, leading to an aggregate sample of 152,141 firm-year observations. We use three different measures of firms cash: pure cash, which is made of cash, cheques and bank deposits divided by total assets (net of pure cash); cash equivalents, which are made up of other short term marketable securities divided by total assets (net of cash equivalents); and total cash, which is the sum of pure cash and cash equivalents divided by total assets (net of pure cash and cash equivalents), as in Opler et al. (1999). As far as the potential cash determinant variables are concerned, we measure size as the natural logarithm of total assets (as used by Opler et al., 1999), where total assets are CPI adjusted at the 2005 price level. A company s risk is proxied by the cash flow 1 The aim is to eliminate firm-year observations registering large jumps in their business fundamentals (typically indicative of major corporate events). volatility, calculated by the standard deviation of cash flows over average total assets, as in Ozkan and Ozkan (2004). The effective tax rate is computed as taxes effectively paid divided by EBT. 2 As it is impossible to adopt the usual market to book ratio in our sample of private firms, growth opportunities are measured as the yearly growth rate of a firm s sales. 3 The financing deficit is the total external financing a company gets in 1 year (either equity or debt) and is proxied by capital expenditures plus dividend payments minus the cash flows generated in that year, as similarly computed by Shyam-Sunder and Myers (1999), 4 and is then divided by total assets. A dividend dummy is a variable set equal to one if the company paid a dividend and equal to zero otherwise. The duration of the cash conversion cycle (in days) is obtained from the inventory (raw material, work-in-progress and finished goods) conversion period plus the receivable collection period minus the payment period for the accounts payable, as in Kim et al. (1998). Financial leverage is financial debt over total assets; bank debt is the ratio of total bank borrowings to total debt, as in Ozkan and Ozkan (2004); net working capital is current assets (net of cash holdings) minus current liabilities divided by total assets; the costs of research and development is scaled on sales and should proxy higher financial distress costs (Opler and Titman, 1994); difference in cash is next s year change in cash holdings and should capture the transitory cash (as in Opler et al., 1999). In order to reduce the possible influence of extreme observations, all variables have been winsorized at the 1% level on both tails of the distributions as in Dittmar and Mahrt-Smith (2007). 5 We also include year dummies and use mean-industry-adjusted variables, 6 as macroeconomic and industry uncertainties are found to have an influence on companies cash holdings (Baum et al., 2006a,b). With regards to the methodology, we allow cash holdings to not adjust immediately to changes in the cash holdings explanatory variables. That is, we allow for some sort of adjustment process to take place, justified by the existence of transaction and adjustment costs. Following Ozkan and Ozkan (2004), we first assume that each ith company has an optimal cash level at year t, function of the aforementioned explanatory variables x k, and an error term l, i.e.: Cash it ¼ a þ X k b k x kit þ l it : If companies do not adjust immediately to their optimal cash levels, the difference between the actual cash and its previous year s level is given by a proportion of the difference between optimal cash and the previous year s cash holdings: Cash it Cash it 1 ¼ kðcash it Cash it 1Þ; where k represents the proportion of the adjustment to the optimal level, ranging between 0 and 1. A value of k equal to 1 indicates that the company adjusts immediately, while a value equal to 0 means that the costs of adjustment are so high that it is inefficient for 2 Negative tax rates have been replaced by zero tax rates. 3 By deriving the yearly implicit growth rate from the observed year till the end of the studied period for which firm-year observations are available. Sales growth as a proxy for a firm s growth is adopted also by Deloof (2001) and Pal and Ferrando (2010). 4 The decomposition does not include the increase in net working capital to avoid multicollinearity problems with the net working capital variable. 5 Only two variables (the effective tax rate and the length of cash conversion cycle) have been winsorized at the 5% level, due to the presence of many extreme and unsound values. 6 Industries in our sample aggregate companies with the same NACE code at the two digit level. The statistical classification of economic activities in the European Community is commonly referred to as NACE (in French: Nomenclature statistique des activités économiques dans la Communauté Européenne ), which is a European industry standard classification system consisting of a six digit code. ð1þ ð2þ

30 M. Bigelli, J. Sánchez-Vidal / Journal of Banking & Finance 36 (2012) 26 35 the company to change its cash level. Since we have continuous data for the 10 year sample period for most of the observations, we can model the cash behavior getting the estimates from the following dynamic panel data specification: Cash it ¼ d 0 þ d 1 Cash it 1 þ d 2 Size it þ d 3 Risk it þ d 4 Eff: tax rate it þ d 5 Growth opportunities it 0.110 0.105 0.100 Total cash Pure cash þ d 6 Financing deficit it þ d 7 Payout dummy it þ d 8 Length of cash conversion cycle it þ d 9 Bank debt it þ d 10 Net working capital it þ d 11 Costs of R & D it þ d 12 Difference in cash it þ g i þ u t þ e it ð3þ Cash/assets 0.095 0.090 where 0.085 d 0 ¼ ak; d 1 ¼ 1 k; d k ¼ kb k ; and e it ¼ kl it : The model (3) results from substituting Eq. (1) into Eq. (2), and then adding both g i, the unobservable time-invariant characteristics of each company which could affect the level of cash holdings, and u t, the year dummy variables reflecting the influence of macroeconomic variables that are common to all firms in a given year and could affect their level of cash holdings. 7 As we have included the lagged dependent variable in this adjustment model, Cash it 1 will be correlated with the g i term that does not vary through time, and therefore OLS parameters will be inconsistent. Another source of bias may arise from possible endogeneity problems if shocks that affect companies cash holdings also affect some of the regressors, as could happen with bank debt, risk, growth opportunities, financing deficit, etc. For these reasons, we estimate the model by using instrumental variable estimators and specifically applying the General Method of Moments (GMM), developed by Arellano and Bond (1991), which obtains consistent parameter estimates by using instruments that can be obtained from the orthogonality conditions existing between the lagged values of the variables and the disturbance terms. When using lagged variables as instruments, the methodology assumes that there is no second-order serial correlation. For this reason we include the Arellano and Bond test for the absence of secondorder serial correlation. Likewise, we test for the absence of correlation between the instruments and the error term with the test of overidentifying restrictions introduced in the context of GMM by Hansen (1982). 4. Results 4.1. Evolution of cash holdings For all 17,165 firms in the sample, Fig. 1 shows the evolution of the average proportion of total assets invested in total cash and pure cash along the sample years. The difference between the two lines reported for total cash and pure cash represents the amount of cash equivalents. The figure indicates that there has been a moderate upward trend in the amount of cash held by Italian private firms in the 1996 2005 period, in line with the increase found by Bates et al. (2009) for US public firms. In fact, average cash holdings represent about 10.6% of total assets in 2005 compared to a lower 9.8% in 1996, and the increase is almost entirely due to an increase of pure cash rather than of cash equivalents. Table 1 reports the univariate statistics for the main variables used in the analysis over the whole sample period. Total cash balances for Italian private companies average 10% of total assets (net 7 This model represents the more complete specification, as we will also use a model considering only firm-specific variables. 0.080 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Years Fig. 1. Evolution of cash holdings. The figure reports the evolution of total cash and pure cash along the sample years (1996 2005) for the 17,165 firms and 152,141 firm-year observations. The values indicated represent the year average of the two cash definitions as percentages of total assets. Pure cash is made up of cash, cheques and bank deposits divided by total assets (net of pure cash ); total cash is the sum of pure cash and cash equivalents divided by total assets (net of pure cash and cash equivalents). of cash), similar to the 8 10.5% level found for US public companies (Kim et al., 1998; Opler et al., 1999; Foley et al., 2007; Mello et al., 2008), almost at the same 9.9% and 10% levels respectively found for UK listed firms (Ozkan and Ozkan, 2004) and Belgian large firms in business groups (Deloof, 2001). The median cash holdings of Italian private firms equals 3.3% of total assets and is below median values found in EMU listed firms (9.1%) by Ferreira and Vilela (2004), in UK listed firms (5.9%) by Ozkan and Ozkan (2004), in Belgian large firms (4.6%) by Deloof (2001) and in Dutch large firms (4.42%) by Bruinshoofd and Kool (2004). Median values are also below the mean in all single industries (Table 2). If we exclude the fishing industry (only seven firms in our sample), more cash holdings seem to be held by firms in real estate, renting and business activities (where the mean and median values are 13.3% and 4.8% respectively), and in transport, storage and communication (where the mean and median values are 12% and 4.7% respectively). 4.2. Small private firms versus large private firms Size should be an important variable in determining the level of financial constraints, the age of the firm, the risk of the firm: all factors which should play an important role in determining the level of cash holdings, especially for private firms which have no access to capital markets. We therefore contrast small and big firms, trying to look for different characteristics and different levels of cash holdings. We rank our sample in size-deciles and we define the observations in the first decile as small firms and those in the last decile as large firms. Results for different firms characteristics and levels of cash holdings are reported in Table 3. As it could be expected, small firms are significantly younger (by almost 8.5 years) and riskier compared to private large firms. They also seem to have a significantly higher effective tax rate that could also be due to fewer opportunities to optimize taxation internationally or within a group. More surprisingly, small firms show significantly fewer growth opportunities. However, being measured by the firm sales yearly growth rate, it means that small

M. Bigelli, J. Sánchez-Vidal / Journal of Banking & Finance 36 (2012) 26 35 31 Table 1 Univariate statistics. Mean Median 25th perc 75th perc St. dev. Total cash/assets 0.100 0.033 0.006 0.114 0.170 Pure cash/assets 0.084 0.028 0.005 0.098 0.139 Cash equivalents/assets 0.010 0.000 0.000 0.000 0.040 Age (years) 24.251 22.000 15.000 30.000 13.361 Size Ln (assets) 9.048 8.950 8.395 9.590 0.962 Risk cash flow volatility 0.045 0.036 0.021 0.058 0.036 Effective tax rate 0.568 0.574 0.420 0.824 0.303 Growth opp. sales yearly growth rate 0.073 0.045 0.010 0.120 0.142 Financing deficit/assets 0.077 0.017 0.039 0.122 0.201 Dividend dummy 0.561 1.000 0.000 1.000 0.496 Length of cash conv. cycle 135.965 135.002 25.360 259.062 145.945 Financial leverage 0.287 0.279 0.113 0.434 0.204 Bank debt/total debt 0.282 0.272 0.054 0.461 0.228 Net working capital/assets 0.071 0.069 0.050 0.198 0.200 Costs of R&D/sales 0.024 0.000 0.000 0.000 0.393 The table reports univariate statistics of the main variables for the 152,141 firm-year observations in the 1996 2005 period. Total cash is the sum of pure cash and cash equivalents divided by total assets (net of pure cash and cash equivalents); pure cash is made up of cash, cheques and bank deposits divided by total assets (net of pure cash); cash equivalents comprise other short term marketable securities divided by total assets (net of cash equivalents); age is the number of the firm s years of life; size is measured as the natural logarithm of total assets (CPI adjusted at the 2005 price level); company risk is proxied by cash flow volatility, measured as the standard deviation of cash flow over average total assets; the effective tax rate is computed as taxes over EBT; growth opportunities is proxied by the sales yearly growth rate; financing deficit is the total external financing a company gets in 1 year (either equity or debt) computed as investments plus dividend payments minus cash flows generated in that year divided by total assets; length of cash conversion cycle (in days) is given by the inventory (raw material, work-in-progress and finished goods) conversion period plus the receivable collection period minus the payment period for the accounts payable; financial leverage is measured by financial debt over total assets; bank debt is the ratio of bank borrowing over total debt; net working capital is current assets net of cash holdings minus current liabilities divided by total assets; costs of R&D/sales is the fraction of sales invested in Research and Development. Table 2 Cash holdings by industry. Industry Number of obs. Percentage of observations in each industry (%) Mean Median Agriculture, hunting and forestry 197 1.148 0.088 0.033 Fishing 8 0.048 0.219 0.136 Mining and quarrying 80 0.465 0.071 0.017 Manufacturing 8459 49.281 0.099 0.033 Construction 1075 6.264 0.074 0.023 Wholesale and retail trade 5100 29.709 0.097 0.030 Hotels and restaurants 164 0.954 0.115 0.029 Transport, storage and communication 800 4.664 0.120 0.047 Real estate, renting and business activities 808 4.706 0.133 0.048 Others: defense, education, other services act., etc. 474 2.761 0.118 0.045 Whole sample 17,165 100.000 0.100 0.033 Industry mean and median values of total cash/assets for the 1996 2005 period. Industries are distinguished by a different NACE code at the 1 digit level. NACE (in French: Nomenclature statistique des activités économiques dans la Communauté Européenne ) is the EU industry standard classification system consisting of a six digit code statistical classification of economic activities. Table 3 Characteristics of small versus large firms. Small firms (1st decile) Large firms (10th decile) Mean diff. t-value Mean Median Mean Median Age (years) 20.248 19.000 27.818 24.000 7.570 48.039 a Risk cash flow volatility 0.050 0.040 0.046 0.035 0.004 8.799 a Effective tax rate 0.609 0.631 0.498 0.494 0.111 31.520 a Growth opportunities 0.067 0.046 0.076 0.053 0.009 6.752 a Financing deficit/assets 0.082 0.010 0.071 0.014 0.011 7.784 a Length of cash conv. cycle 154.368 180.489 111.179 101.246 43.189 25.258 a Financial leverage 0.201 0.167 0.371 0.377 0.170 71.289 a Bank debt/total debt 0.201 0.153 0.297 0.283 0.097 37.348 a Net working capital/assets 0.054 0.053 0.074 0.070 0.020 8.234 a Costs of R&D/sales 0.007 0.000 0.025 0.000 0.017 4.110 a Total cash/assets 0.135 0.053 0.075 0.021 0.059 29.595 a Pure cash/assets 0.120 0.048 0.056 0.018 0.064 39.172 a Cash equivalents/assets 0.008 0.000 0.012 0.000 0.004 9.494 a Analysis of the main characteristics of small and large firms, respectively defined as those in the first and tenth decile for the variable size (natural logarithm of total assets, price adjusted). The table reports mean and median values for the main variables and for total cash holdings, pure cash and cash equivalents in small and large firms. Mean differences and their respective t-values are reported in the last two columns. a Indicates significance at the 1% level.

32 M. Bigelli, J. Sánchez-Vidal / Journal of Banking & Finance 36 (2012) 26 35 firms grow less than large firms. Such a result may sound counterintuitive but we must consider that our sample includes all industries. As a matter of fact, 30% of the whole sample is made up of wholesale and retail trade companies (see Table 2), and small companies in this industry have suffered strong competition from large players in the last decade. As far as the financial characteristics of small firms is concerned, they appear to have a significantly larger financing deficit, a longer cash conversion cycle, a lower financial leverage, a lower proportion of bank debt on total debt, a lower amount of net working capital and lower costs of R&D. The lower financial leverage and bank debt could be due to the higher riskiness and lower credit worthiness of small firms while the longer cash conversion cycle could be due to their lower contracting power in payables and receivables terms of payment, compared with the biggest private firms. When we look at differences in cash holdings, results are as expected. In fact, small firms, being more risky and more financially constrained, tend to hold significantly more total cash than large firms (13.2% versus 7.5% of total assets) as expected under the trade-off model and hypothesis 1. Such a sharp difference is due entirely to pure cash balances (11.8% versus 5.6%), as cash equivalents represent only a minor fraction of total cash holdings and tend to be held more by large firms than small firms (1.2% versus 0.7%). As in Opler et al. (1999), Bruinshoodf and Kool (2004), and Harford et al. (2008) the cash-substitutes arguments find empirical support. In fact, cash can be considered as negative debt and small firms report significantly lower leverages and bank debts. Small firms holding more cash have also significantly lower fractions of working capital, a typical cash-substitute. 4.3. Determinants of cash holdings 8 Since the panel is incomplete and we use lags as instruments, the number of firms used in the regressions is slightly below the 17,165 firms of the sample. 9 The pearson correlation coefficient is equal to 0.081, the highest value and the most significant between all the explanatory variables. Table 4 shows the GMM regressions on the determinants of cash holdings. All the estimations have been carried out using the two-stage GMM estimator. The two-stage employs the residuals on the one-stage estimation to construct an asymptotically weighted optimum matrix and it is more efficient than the onestage if we assume that perturbances will show heteroskedasticity for relatively extensive sample data (Blundell and Bond, 1998), such as ours. We treat as endogenous the variables for which the Hansen test validates their instrumentation. Models 1 and 2 present only firm-specific variables, while model 3 includes year dummies and models 4 and 5 are run on industry-adjusted variables. The absence of any second-order serial correlation in all the three models confirms the consistency of our estimations, while the acceptance of the Hansen test validates the correctness of the number of selected instrumental variables, as the instruments and the residuals are independent. 8 In all five regressions the lagged dependent variable is not significant, probably because of a quite heterogeneous behavior amongst firms with respect to their adjustment process to the optimal cash level. In our first model we can observe that most of the expected relationships under the trade-off model find empirical support: firms characterized by smaller size, riskier cash flows and lower effective tax rates hold significantly higher amounts of cash reserves. Similar results for the firm s size and risk have been found by Opler et al. (1999). We find that growth opportunities are not significant, maybe because of a high positive correlation with the financing deficit variable. 9 As this last variable presents a significant negative sign in the regression, it could have transposed this negativity to the growth opportunities variable making it lose part of the expected positive influence. Our evidence on private firms strongly supports the relationship predicted by the financing hierarchy theory. In fact, firms with higher financing deficits hold significantly lower amounts of cash. 10 Moreover, the significant sign of the payout dummy also confirms the expectation that private firms that pay dividends hold more cash, differently from what is found for US public firms (Opler et al., 1999; Harford et al., 2008). This difference could not only be due to private firms not suffering from agency costs of free cash flows that characterize US public companies, but also to institutional differences in dividend payments. In fact, US listed firms pay quarterly dividends while Italian private (and public) companies tend to pay dividends once a year, typically in the second quarter, after the approval of financial statements. It follows that companies that regularly distribute yearly dividends are likely to end the fiscal year (in December) with more cash reserves in order to meet the dividend payment to be disbursed in a few month s time. Finally, firms with a longer cash conversion cycle hold significantly more cash as theoretically expected. 11 The signs and significance of the two cash-substitute variables provide support for the predicted relations. In fact, firms having a greater proportion of bank debt and more net working capital hold significantly less cash, confirming the general wisdom that cash can be considered as negative debt, and net working capital can be considered as a valid cash substitute, as also found by Ozkan and Ozkan (2004) on UK public firms and Ferreira and Vilela (2004) on EU listed firms. The fraction of sales invested in R&D does not seem to greatly affect the amount of cash holdings, probably because they are not so relevant in most private small firms. Model 2 adds to model 1 the variable difference in cash (computed as next year s change in cash holdings) which should capture the part of cash holding that is transitory kept. Similarly to Opler et al. (1999), the variable reports an insignificant negative coefficient and shows little impact on the coefficients of the other variables. Model 3 adds year dummies to explore the possibility that companies cash balances can be influenced by some macroeconomic variables or by the economic cycle, though their effects should be at least partially incorporated into the financing deficit variable we are using. 12 When we introduce year dummies in our regression model we find that some years are statistically significant, while some variables lose their previous significance, specifically the financing deficit and the length of cash conversion cycle. This suggests that the influence of such variables on cash holdings could also depend on macroeconomic variables and the business cycle. In models 4 and 5 we explore to what extent industry factors can influence the level of companies cash balances. We therefore use industry-adjusted variables, obtained as differences between the values of the firm-year variables and the values of the industry- 10 With respect to this variable, we explore the fact that cash can also be affected by the previous year s financing deficit, and we find similar (unreported) results. We also disaggregate the financing deficit into operating cash flows, investments and distributed dividends and find that all three components exhibit the sign expected under the financial hierarchy theory (positive for the cash flows and negative for the investments and dividends). However, none of them show statistical significance and their disaggregation makes some other variables lose their statistical significance (the payout dummy and the length of the cash conversion cycle). We believe this is due to the high correlation between the investment variable and the cash flow variable and, above all, to the high positive correlation between the dividends variable and the payout dummy variable. For this reason, we decide to leave the financing deficit variable without disaggregating it in its components. 11 The coefficient for this variable exhibits very low figures (as in Kim et al., 1998), because of the high values of this independent variable. 12 As we include a considerable number of explanatory variables, in order to check for a potential problem of multicollinearity we check that the critical level of 5 for the VIF-value (variance inflation factor) is not surpassed in any model.

M. Bigelli, J. Sánchez-Vidal / Journal of Banking & Finance 36 (2012) 26 35 33 Table 4 GMM cash holdings regressions: dynamic panel data estimation results. Variables Models 1 2 3 4 5 Cash holdings t 1 0.071 0.061 0.249 0.222 0.223 Size 0.008 a 0.008 a 0.005 b 0.002 0.002 Risk 0.394 a 0.414 a 0.286 a 0.262 a 0.263 a Effective tax rate 0.016 a 0.016 a 0.011 a 0.013 a 0.013 a Growth opportunities 0.012 0.003 0.050 0.056 0.060 Financing deficit/assets 0.169 a 0.166 b 0.122 0.178 0.175 Payout dummy 0.027 a 0.026 a 0.019 a 0.020 a 0.020 a Length of c. c. c. 0.000 a 0.000 a 0.000 0.000 0.000 Bank debt/total debt 0.244 a 0.245 a 0.169 a 0.156 a 0.157 a Net working capital/assets 0.267 a 0.263 a 0.201 a 0.190 a 0.191 a Costs of R&D/sales 0.006 0.061 0.261 0.054 0.060 Difference in cash 9.506 Year 1999 dummy 0.007 a 0.000 Year 2000 dummy 0.001 0.002 Year 2001 dummy 0.002 0.002 Year 2002 dummy 0.001 0.003 Year 2003 dummy 0.008 a 0.002 Year 2004 dummy 0.014 a 0.004 c Number of firms 17,043 16,987 17,043 17,043 17,043 m 2 1.34 (0.180) 1.35 (0.183) 0.5 (0.864) 0.25 (0.799) 0.26 (0.799) Hansen test 17.76 (0.123) 16.94 (0.110) 15.36 (0.222) 13.15 (0.358) 13.05 (0.366) This table presents GMM regressions predicting cash holdings. The sample period runs from 1996 till 2005 although the available number of years for each company changes across firms. Models 1 and 2 present the 2-stage GMM regressions on the major expected determinants of cash holdings. Model 3 includes year dummies while models 4 and 5 run the regression with mean industry-adjusted variables. Size is the natural logarithm of real total assets; company risk is proxied by cash flow volatility, measured as the standard deviation of cash flow over average total assets; effective tax rate is taxes over EBT; growth opportunities is the yearly growth rate of a firm s sales; financing deficit is investments plus dividend payments minus operating cash flows; dividend dummy is a variable equal to one if the company paid a dividend in the year and equal to zero otherwise; length of cash conversion cycle (in days) is given by the inventory (raw material, work-in-progress and finished goods) conversion period plus the receivable collection period minus the payment period for the accounts payable; bank debt is the ratio of bank borrowing over total debt; net working capital is current assets net of cash holdings minus current liabilities; costs of R&D/sales is the fraction of sales invested in Research and Development; difference in cash is the change in cash over net assets from year t to year t + 1. Year dummies are set to one in the observation year. We have eliminated year 1998 and year 2005 dummies in model 2 to avoid multicollinearity. The intercept coefficient is not included. m 2 is a test for second-order serial autocorrelation in the residuals. Hansen test is distributed as Chi-square under the null of instrument validity. Significance in brackets. a Indicates significance at the 1% level. b Indicates significance at the 5% level. c Indicates significance at the 10% level. Table 5 High-cash firms and new investments, dividend payments, debt repayment and future profitability. High-cash firms Rest of the sample Difference (t-value) t + 1 year investments 0.001 0.000 0.002 2.784 a t + 2 year investments 0.005 0.002 0.006 8.483 a Following 4 years average investments 0.002 0.001 0.002 5.810 a t + 1 year dividends 0.010 0.003 0.014 59.118 a t + 2 year dividends 0.010 0.003 0.013 48.857 a Following 4 years average dividends 0.010 0.003 0.012 54.054 a t + 1 year debt repayment 0.036 0.048 0.011 13.345 a t + 2 year debt repayment 0.036 0.042 0.006 5.978 a Following 4 years average debt rep. 0.038 0.046 0.007 15.304 a t + 1 year profitability 0.018 0.006 0.024 58.149 a t + 2 year profitability 0.014 0.007 0.021 47.079 a Following 4 years average profitability 0.015 0.005 0.020 44.181 a The table reports a comparison of future investments, dividend payments, debt repayment and future profitability between high-cash firms (first quartile of firms for cash holdings) and the rest of the sample. Future investments and dividends are scaled by total assets; debt repayment is the decrease of financial debt from the previous year and is also scaled by total assets; profitability is the operating income on total assets. Variables are measured within 1 year, 2 years and as an average of the 4 years after the cashvariable measurement-year. All variables are industry-adjusted. a Indicates significance at the 1% level. year means for those variables. 13 When we use industry-adjusted variables, most of them keep their statistical significance, which could mean that although the firm s industry may play a role, firm-specific variables continue to be relevant in terms of cash-holdings explanation. However, industry factors seem to be relevant in determining the explanatory effects of the firm s size, financing deficit and length of cash conversion cycle, as these variables are no 13 The industry-adjusted variables are obtained using industry classification at the two-digit NACE level. more significant, when adjusted for the industry levels. Years dummies (1999 and 2003 and partially 2004) also lose their statistical significance, indicating that macroeconomic factors affect the cash level of firms but also the cash level of the whole industries. 4.4. How is cash spent? We try to analyze the use of cash by cash-rich firms to find out if the high cash reserves are used to finance new investments, distribute dividends, repay debt in the following years and if they