Institutional Ownership and Firm Cash Holdings

Institutional Ownership and Firm Cash Holdings Christine Brown Yangyang Chen Chander Shekhar May 2011 Corresponding author. Brown (christine.brown@monash.edu) and Chen (yangyang.chen@monash.edu) are at Department of Accounting and Finance, Monash University, VIC, Australia. Shekhar (c.shekhar@unimelb.edu.au) is at Department of Finance, The University of Melbourne, VIC, Australia.

Institutional Ownership and Firm Cash Holdings Abstract The precautionary theory of firm cash holdings argues that firms hold cash to protect themselves against adverse cash flow shocks or external financial constraints that might force them to default on payments or forgo valuable investment opportunities. The aggressive trading of a group of institutional investors magnifies the firm s stock price volatility and makes external financing costly. This may induce the firm s precautionary demand for cash. Consistent with the hypothesis, we find that firms with greater institutional ownership have more cash holdings. The relationship is mainly driven by the ownership of short-term institutions (i.e., institutions that trade frequently for short-term trading profits). The ownership of long-term institutions (i.e., institutions that trade infrequently) is negatively associated with firm cash holdings. Further tests show that the effect of both types of institutional ownership on firm cash holdings is more significant for growth firms that rely more on external financing and suffer more from cash shortages. JEL Classification: G32 Keywords: Cash holdings; Precautionary savings theory; Institutional Ownership. 1

1. Introduction The importance of cash holdings in a firm s overall financial health is reflected in the business view that cash is king. A firm could have a large amount of physical assets or accounts receivable on its balance sheet, but may still run out of cash, resulting in technical bankruptcy or loss of valuable growth opportunities. Therefore, firms tend to stockpile cash to protect against temporary cash shortages in making payments or investments. Previous studies (e.g. Opler et al. (1999), Harford, Mansi, and Maxwell (2008)) document various determinants of firm cash holdings, including growth status, business risk, financial policy and agency problems. However, to date there has been no comprehensive analysis of the role of ownership structure, especially institutional ownership, on firm cash holdings. Institutional investor ownership has grown substantially in recent decades, attracting attention from financial economists. A considerable body of research has explored the influence of institutional ownership on various aspects of firm operations, including payout policy (e.g., Grinstein and Michaely (2005)), mergers and acquisition (e.g., Chen, Harford, and Li (2007)), and executive compensation (e.g., Hartzell and Starks (2003)). This study examines how institutional investors impact the propensity of firms to stockpile cash in the US. More specifically, we propose and test two competing hypotheses on the relationship between institutional ownership and firm cash holdings. The trading hypothesis argues that a group of institutional investors has trading strategies based on the likelihood of short-term trading profits. Their frequent and 2

large-scale trading generates short-term price pressure for the stock, resulting in volatile stock prices. The herding behavior of some active trading institutions magnifies stock market volatility further. Bushee and Noe (2000) find that the ownership of institutions with a short-term focus and high portfolio turnover is associated with more volatile stock returns. They also find that the ownership of institutions that provide long-term capital to the firm has the opposite effect, perhaps because these institutions are less likely to engage in aggressive trading activities. High stock market uncertainty makes the firm's external financing, especially equity financing, more costly, because firms will face higher direct costs (Corwin (2003), Eckbo, Masulis, and Norli (2007)) and more uncertainty in equity issuance. Therefore, the ownership of aggressive trading institutions imposes additional external financial constraints on the firm, inducing its precautionary demand for cash. The trading hypothesis thus predicts a positive relationship between institutional ownership and firm cash holdings. The monitoring hypothesis states that institutional investors, by virtue of their relatively large stock holdings and expertise in evaluating firm operations, are better able to monitor management than individual investors. But this observation can have two quite opposite effects on the level of a firm s cash holdings. As noted by Jensen (1986), entrenched managers would rather retain cash for their own interests than increase payouts to shareholders when the firm has poor investment opportunities. If institutional monitoring helps to reduce this free cash flow problem, then a negative relation between institutional ownership and firm cash 3

holdings is expected. We call this the Type 1 monitoring hypothesis. On the other hand, institutional monitoring prevents managers from deploying cash on value-destroying capital expenditure and acquisitions. Dittmar and Mahrt-Smith (2007) and Harford, Mansi, and Maxwell (2008) find that firms with weaker corporate governance have smaller cash reserves. These firms tend to cut payouts but spend the excess cash quickly in capital expenditures and acquisitions, resulting in a lower level of cash holdings, and a positive association between institutional ownership and firm cash holdings may result. We call this the Type 2 monitoring hypothesis. Employing a sample of nonfinancial and nonutility Compustat firms from 1981 to 2007, we first show that the level of firm cash holdings increases with a rise in total institutional ownership, defined as the percentage of shares owned by all the institutional investors. However, institutional investors have different investment horizons and trading strategies that may have varying impact on firm operations. Recognizing the heterogeneity among institutional investors, we classify them into short-term and long-term using the method proposed by Yan and Zhang (2009). The classification is based on the institution's aggregate portfolio turnover in the past four quarters. The regression results show that short-term institutional ownership is positively related, while long-term institutional ownership is negatively related, to firm cash holdings. In addition, after controlling for short-term institutional ownership, the coefficient of total institutional ownership becomes negative, suggesting that the positive relation between total institutional ownership and cash holdings is purely 4

driven by short-term institutions. The results are robust to the omitted variables problem, reverse causality, simultaneity, and spurious regression due to a time-series trend. Further, the same conclusions are drawn using Bushee s (1998) institutional investor classification data. Short-term institutions focus on short-term trading profits and have high portfolio turnover. This group of institutional investors engages more in aggressive trading activities than long-term institutions, which provide stable capital to the firm and earn profits (e.g., dividends and capital appreciation from long-term holdings. Short-term institutions are also less likely to monitor management because they have insufficient monitoring capacity, suffering higher costs and enjoying less benefit from monitoring. Therefore, the results support the trading hypothesis while providing evidence against both types of monitoring hypothesis. The results for long-term institutional ownership support both the trading hypothesis and Type 1 monitoring hypothesis. Since long-term institutions provide stable capital to the firm, their ownership stabilizes the firm s stock price and reduces external financial constraints, resulting in less precautionary demand for cash. Furthermore, long-term institutions are more likely to engage in monitoring activities. Their presence reduces the free cash flow of the firm, resulting in the firm holding less cash. Last, we provide further evidence by examining the differing effect of institutional ownership on growth and mature firms. Since growth firms are normally unprofitable, they rely more on external financing, especially equity financing, than mature firms 5

(Frank and Goyal (2003)). They also have more firm value allocated to future investment opportunities and thus suffer more from external financial constraints. Therefore, the trading hypothesis predicts that growth firms have a precautionary demand for cash which is more sensitive to changes in institutional ownership than that of mature firms. The results support this prediction by showing that the effects of both short-term and long-term institutional ownership are more prominent for growth firms than for mature firms. However, the results are inconsistent with the Type 1 monitoring hypothesis that institutional ownership mitigates the firm s free cash flow problem and thus reduces its cash holdings. Mature firms rather than growth firms suffer more from free cash flow problems. If the Type 1 monitoring hypothesis were true, more significant results should be observed for mature firms, not growth firms. The contributions of this study to the literature are threefold. First, the study is the first to comprehensively investigate the effect of institutional ownership on firm cash holdings. Harford, Mansi, and Maxwell (2008) employ institutional ownership as a proxy for corporate governance strength, in parallel with other measures such as the GIndex and board independence, and document a positive but mostly insignificant relation between it and firm cash holdings. Employing a much larger sample, this study documents a robust relation between different types of institutional ownership and firm cash holdings. The results establish an important determinant of firm cash holdings institutional ownership structure which has been ignored by prior research and suggests a new angle for understanding the firm's financial policy. Second, and deviating from models that study the relation between institutional 6

ownership and firm cash holdings from an agency point of view, this study proposes an explanation based on institutional investors' trading behavior in the stock market and finds supporting evidence. The findings suggest that there is no single role attributable to institutional investors. Besides monitoring, their trading activities may also affect firm operations or even drive some of the observed patterns (e.g., changes in firm cash holdings) through changing a firm's stock price behavior. 1 Third, the study highlights the importance of differentiating institutional investors based on their characteristics. Institutional investors are not an homogeneous group. They have different investment objectives, trading style, legal restrictions, and competitive pressures that make them behave differently. By classifying institutional investors based on their investment horizon this study finds that different groups of institutions have opposite impacts on firm cash holdings. The results caution against practices that treat institutional investors as a homogeneous group as their impact on firm operations is comprehensive and varies among institutions with different characteristics. Oversimplifying the analysis may cause important information to be missed. The rest of the paper is organized as follows. Section 2 reviews theories on firm cash holdings and develops the hypotheses. Section 3 describes data sources, variable construction and summary statistics. Section 4 presents univariate and multivariate results for the empirical tests. Section 5 conducts robustness checks. Section 6 provides further evidence on the hypothesis and Section 7 concludes. 1 McLean (2011) reports that firms increasingly issue shares to meet precautionary motive for cash savings. Our results are complimentary to his as they suggest that firms hold more cash if they are unable to issue shares, which itself is a result of ownership by short-term institutional investors. 7

2. Hypotheses 2.1. Theories of Firm Cash Holdings In a perfect capital market, holdings of cash or other liquid assets are irrelevant to firm value (Modigliani and Miller (1958)). When cash flow is not enough for payments or investments, firms can raise funds by liquidating real assets and/or raising external capital without any costs. Therefore, cash holdings in a perfect market are purely random. However, in the real world, cash holdings are not random. Firms have incentives to stockpile cash and maintain their cash holdings at a certain level. There is no shortage of theories to explain firms propensity to hold cash; in this section we review the main ones. The early models of Baumol (1952) and Miller and Orr (1966) focus on the transaction costs incurred in raising funds through asset sales or security issuances. Firms have an incentive to hold cash rather than seeking more expensive alternative funding sources..since there are economies of scale in transaction costs, the model predicts that large firms hold less cash, which is supported by the empirical findings of Mulligan (1997). Rather than holding cash to reduce the transaction costs incurred in raising funds, Keynes (1936) argues that firms hold cash to protect themselves against adverse cash flow shocks and external financing constraints that might force them to default on payments or forgo valuable investment opportunities. In support of the theory, Opler et al. (1999) document that firms with more growth opportunities and riskier cash 8

flows hold relatively more cash, while firms with better access to the capital markets, such as large firms and those with high credit ratings, maintain a lower level of cash holdings. Bates, Kahle, and Stulz (2009) observe a substantial increase in firm cash holdings in recent decades, which they attribute to precautionary motives. During this period, firms have riskier cash flows, hold fewer inventories and receivables, and are increasingly R&D intensive. They have more precautionary demand for cash and thus maintain higher cash reserves. Further, Duchin (2010) finds that multi-division firms hold significantly less cash than standalone firms because they are diversified in their cash flows and investment opportunities. The diversification effects are stronger in financially constrained firms. Almeida, Campello, and Weisbach (2004) find that financially constrained firms rely more on internal cash flow in building their cash holdings (i.e., positive cash flow sensitivity of cash) while non-constrained firms do not. Furthermore financially constrained firms increase their cash holdings in response to increases in cash flow volatility, while unconstrained firms do not exhibit such a tendency (Han and Qiu, 2007). Consistent with the view that financial development eases financial constraints by improving a firm's access to lower cost external financing, in a cross-country study, Khurana, Martin, and Pereira (2006) show that the cash flow sensitivity of cash is negatively related to a country's level of financial development.. While Riddick and Whited (2009) question existing results for the cash flow sensitivity of cash as the results may be affected by measurement error in Tobin s Q, they neverthelessdocument that firms hold higher precautionary cash balances when 9

external finance is costly or income uncertainty is high. Agency theory of firm cash holdings as initiated by Jensen (1986) states that entrenched managers would rather retain cash for their own interests than increase payouts to shareholders when the firm has poor investment opportunities. Dittmar, Mahrt-Smith, and Servaes (2003) find cross-country evidence that firms in countries where shareholders rights are not well protected hold substantially larger cash balances than firms in countries with good shareholder protection. Pinkowitz, Stulz, and Williamson (2006) show that the value of cash holdings is less in countries with poor investor protection, and Dittmar and Mahrt-Smith (2007) show that well-governed U.S. firms enjoy higher valuation of their cash holdings. However, Dittmar and Mahrt-Smith (2007) and Harford, Mansi, and Maxwell (2008) document that firms with weaker corporate governance actually have smaller cash reserves. These firms tend to build excess cash balances but spend cash quickly through capital expenditures and acquisitions, which reduces the profitability and valuation of the firm. The product market theory states that firms may hold cash to gain competitive advantage in the product market. Haushalter, Klasa, and Maxwell (2007) find that firms use more derivatives and hold more cash if they share a large proportion of growth opportunities with rivals, because an inability to fully invest in these opportunities will lead to predatory behavior on the part of rivals and loss of market share. Fresard (2010) shows that large cash reserves lead to systematic future market-share gains at the expense of industry rivals. This effect is stronger when 10

rivals face tighter financing constraints and when interaction between competitors is large. Schroth and Szalay (2010) focus on the pharmaceutical industry and document that firms are more likely to win patent races the more cash and assets they hold and the less cash and assets their rivals hold prior to the race. 2.2. Hypothesis Development Building upon the existing theories of firm cash holdings, in this section we develop two testable hypotheses for the role that institutional investors play in determining the level of firm cash holdings. 2.2.1. The Trading hypothesis A group of institutional investors base their trading strategies on the likelihood of short-term trading profits. The large-scale trading and herding of these institutions magnify stock price volatility. First, institutional investors normally trade in much larger volumes than individual investors. Their trading generates short-term price pressure that results in large movements in stock prices. Porter (1992) finds that the aggressive trading behavior and alleged propensity to engage in large-scale selling when faced with bad news lead to higher stock return volatility around news events. Chan and Lakonishok (1993, 1995) show that packages of trades by institutional investors have sizable price impacts over short horizons, with high turnover institutions producing the largest price impacts. Second, there is evidence that institutional investors are more susceptible to herding behavior than individual 11

investors because of the importance of relative performance and the asymmetry of incentives for fund managers (e.g., Nofsinger and Sias (1999), Sias (2004)). Dennis and Strickland (2002) show that institutional herding contributes, at least in the short term, to stock market volatility. As supporting evidence, Sias (1996) finds that increases in institutional ownership precede increases in stock return volatility. However, the relationship is not robust as Gompers and Metrick (1998) find the opposite. Bushee and Noe (2000) reconcile the conflicting results by showing that institutional investors with different investment strategies have differing impact on stock return volatility. The ownership of institutions with short-term focus and high portfolio turnover is associated with greater stock return volatility; while the ownership of institutions that provide long-term, stable capital is associated with smaller stock return volatility. High stock return volatility increases firms' external financing risk, especially the risk of equity financing, inducing a precautionary motive to hoard cash. If stock prices fluctuate too much, firms will face more uncertainty in future equity financing. In cases where they need external financing but stock prices are temporally low, firms may be forced to give up positive NPV projects to avoid issuing undervalued stock. Moreover, direct costs of issuing equity securities also increase with stock return volatility. There is evidence that underwriting spread rises with a security's total risk, measured by its return volatility around the issuance (Eckbo, Masulis, and Norli (2007)). SEO underpricing is also positively related to price uncertainty (Corwin (2003)). Therefore, equity financing is restricted when stock returns are volatile. 12

Taken together, these arguments lead to the trading hypothesis of the relationship between institutional ownership and firm cash holdings. Trading hypothesis: The aggressive trading activities of institutional investors destabilize stock prices and lead to higher external financing risk. As a consequence, firms have more precautionary demand for cash and thus maintain higher cash reserves. The hypothesis predicts a positive relationship between institutional ownership and firm cash holdings. 2.2.2. The Monitoring Hypothesis Institutional investors are able to monitor management more effectively and efficiently than individual investors. They have expertise in evaluating firm operations and enjoy greater voting power in taking corrective actions when necessary. These investors also enjoy more benefits and incur lower costs from monitoring because of their relatively large stock holdings. There is evidence that institutional ownership is associated with more value-enhancing antitakeover charter amendments (Agrawal and Mandelker (1990)) and that the voting outcome of shareholder proposals is strongly associated with whether the sponsors are institutional investors and the percentage of institutional ownership (Gillan and Starks (2000)). Further, Hartzell and Starks (2003) show that institutional ownership concentration is associated with improvement in management compensation and Chen, Harford, and Li (2007) find that the existence of concentrated ownership by independent long-term institutions yields better post-merger performance for the acquiring firm. 13

On the one hand, institutional monitoring reduces the firm s free cash flow problem. Jensen (1986) states that entrenched managers would rather retain cash for their own interests than increase payouts to shareholders when the firm has poor investment opportunities. Dittmar, Mahrt-Smith, and Servaes (2003) find cross-country evidence that firms in countries where shareholders rights are not well protected hold substantially larger cash balances than firms in countries with good shareholder protection. On the other hand, institutional monitoring prevents managers from deploying cash on value-destroying capital expenditure and acquisitions. Dittmar and Mahrt-Smith (2007) show that poorly-governed U.S. firms suffer from lower valuation of their cash holdings, as these firms dissipate cash quickly in ways that significantly reduce operating performance. Harford, Mansi, and Maxwell (2008) demonstrate that firms with weaker corporate governance structures have smaller cash reserves. These firms tend to choose repurchases instead of increasing dividends to avoid future payout commitments. Any excess cash built up is spent quickly in value-destroying capital expenditures and acquisitions, resulting in a lower level of cash holdings. Institutional monitoring improves the firm's corporate governance, reducing the scope of management discretion. Harford, Mansi, and Maxwell (2008) document a positive but mostly insignificant relationship between institutional ownership and firm cash holdings and offer institutional monitoring as the explanation. Thus institutional ownership can be argued to have a positive or a negative effect on firm cash holdings, which leads to the statement of the following two hypotheses. 14

Type 1 monitoring hypothesis: Institutional monitoring reduces the firm s free cash flow problem through forcing managers to pay out excess cash to shareholders. The hypothesis predicts a negative relationship between institutional ownership and firm cash holdings. Type 2 monitoring hypothesis: Institutional monitoring reduces management discretion in making capital expenditure and acquisitions, which prevents self-interested managers from deploying cash quickly. The hypothesis predicts a positive relationship between institutional ownership and firm cash holdings. 3. Data 3.1. Institutional Ownership Institutional ownership data comes from Thomson Financial's CDA Spectrum database, available on Wharton Research Data Services (WRDS) 2. The 1978 amendment to the Securities and Exchange Act of 1934 requires that all institutional investors with holdings of $100 million or more under management must report their holdings quarterly using Form 13F of the Securities and Exchange Commission (SEC); common stock positions greater than 10,000 shares or $200,000 must be filed. Thomson Financial collects and reports holdings information of each institutional investor for each firm on a quarterly basis since the first quarter of 1980. These institutions include bank trusts, insurance companies, investment companies (mutual funds), independent investment advisors (mainly large brokerage firms), ESOPs, 2 http://wrds.wharton.upenn.edu 15

foundations, university endowments and pension funds (both public and private). The holdings information is based on the entire institution (i.e., the legal entity) rather than its individual divisions. Although institutional investors share some commonalities, they are far from a homogeneous group. Institutional investors are distinguished by their investment objectives, trading style, legal restrictions, and competitive pressures. These differences engender different investment horizons and trading strategies, which are likely to affect firm operations and stock price behavior. To distinguish between types of institutional investors, we adopt the institutional investor classification method from Yan and Zhang (2009), who group institutional investors into short-term and long-term based on their aggregate portfolio turnover. The classification operates as follows. First, we calculate the aggregate purchase and sale for each institution: N k k, t k, i, t i, t k, i, t 1 i, t 1 k, i, t 1 S i, t k, i, t Sk, i, t 1 i 1 CR _ buy S P S P S P (1) N k k, t k, i, t i, t k, i, t 1 i, t 1 k, i, t 1 S i, t k, i, t Sk, i, t 1 i 1 CR _ sell S P S P S P (2) where P i,t-1 and P i,t are the share prices for stock i at the end of quarter t-1 and t, and S k,i,t-1 and S k,i,t are the number of shares of stock i held by investor k at the end of quarter t-1 and t, respectively. We adjust for stock splits and stock dividends using the CRSP price adjustment factor. CR_buy k,t and CR_sell k,t are institution k s aggregate purchase and sale for quarter t, respectively. Institution k s churn rate for quarter t is 16

then defined as: CR kt, min( CR _ buyk, t, CR _ sellk, t ) Nk S P S P i 1 k, i, t i, t k, i, t 1 i, t 1 2 (3) as: Next, we calculate each institution s average churn rate over the past four quarters 3 1 AVG _ CR CR (4) k, t k, t j 4 j 0 Finally, we sort all institutional investors into three groups each quarter based on their average churn rate. The group with the highest average churn rate is labeled as short-term institutional investors, while the group with the lowest average churn rate is labeled as long-term institutional investors. Institutional ownership is defined as the fraction of a firm's shares held by institutional investors. The definition is widely adopted in prior literature (e.g. Gompers and Metrick (2001), Grinstein and Michaely (2005), Chen, Harford, and Li (2007)). We calculate total institutional ownership in the following way: in each quarter, we sum the shares held by all the institutional investors for each firm and then divide the sum by the firm's number of shares outstanding. Short-term and long-term institutional ownership are defined as the fraction of shares held by corresponding institutional investors and are calculated in the same way as total institutional 17

ownership. 3.2. Firm Cash Holdings and Control Variables Firm financial information comes from the merged Compustat/CRSP Fundamental Annual database, available on WRDS. The database covers more than 24,000 active and inactive publicly held firms in the US and Canada from 1950. Most of the firms are traded on the New York Stock Exchange (NYSE), American Stock Exchange (AMEX) or Nasdaq, and the rest are traded on regional US stock exchanges or Canadian stock exchanges. We define firm cash holdings as the proportion of cash and cash equivalents in the firm s total assets. Following Opler et al. (1999) and Bates, Kahle, and Stulz (2009), we include measures of firm characteristics that may affect firm cash holdings as control variables. The definitions of these variables are presented below. Detailed information of variable construction is available in Table 1. Industry cash flow risk - Standard deviation of industry cash flow to assets. Market-to-book ratio - The ratio of market value of assets over book value of assets. Firm size - Natural logarithm of book value of assets. Cash flow - The ratio of operating cash flow over book value of assets. Net working capital - The ratio of working capital net of cash over book value of assets. Capital expenditure - The ratio of capital expenditure over book value of assets. 18

Leverage - The ratio of book value of debt over book value of assets. Dividend dummy - Dummy variable equal to one if the firm pays common dividends and zero otherwise. R&D - The ratio of R&D expenses over sales revenue. Acquisition - The ratio of acquisition expenditure over book value of assets. 3.3. Sample and Summary Statistics We merge financial information with institutional ownership data using six-digit Committee on Uniform Security Identification Procedures (CUSIP-6) as firm ID. Note that financial information has fiscal year frequency, while institutional ownership data has calendar quarter frequency. In merging these two datasets, we match financial information with institutional ownership of the quarter immediately before the fiscal year end to make sure that financial data is set to be always lagged or at least contemporaneous to institutional ownership data. Another problem in the merging process is that a large proportion of observations in Compustat do not have institutional ownership information in CDA Spectrum. Following Gompers and Metrick (2001), if a firm in Compustat is not held by any institution, we set its institutional ownership to zero. To control for the impact of missing values, we include into the regressions a dummy variable equal to one if the firm has institutional ownership information in CDA Spectrum and zero otherwise. Since the institutional investor data traces back to 1980 and four quarters data are needed for the classification, we restrict the sample period to 1981-2007. Following 19

prior studies in firm cash holdings (e.g., Bates, Kahle, and Stulz (2009)), we drop financial firms with SIC code 6000-6999 and utility firms with SIC code 4900-4999. Firms in these industries may hold cash for noneconomic reasons (e.g., capital requirements or regulation), which may distort the analysis if included. We require that firms have positive book assets and sales and that leverage and all types of institutional ownership lie within the closed unit interval. To mitigate the effect of outliers, we winsorize cash flow and net working capital at the lower, and market-to-book ratio at the upper, one-percentile. All the other variables are winsorized at both the upper and lower one-percentile. Dollar amounts are deflated to 2000 dollars. Summary statistics of the variables are exhibited in Table 2. The statistics show that the mean total institutional ownership of the sample firms is 20.1%, among which 8.5% is held by short-term institutions and 4.1% by long-term institutions. Also, 63.8% of Compustat firms have nonmissing institutional ownership data. The sample firms on average hold 16.4% of their total assets as cash and cash equivalents. They also hold 4.6% of total assets as net working capital and spend an amount equivalent to 7.2% and 2.1% of total assets on capital expenditure and acquisition, respectively. Further, firms in the sample have mean cash flow of -4.8% of total assets. Nevertheless, the median cash flow is 5.3%, suggesting that the mean is driven by extreme values. The R&D to sales ratio is similarly skewed, with the mean and median values 20.1% and 0%, respectively. Last, 29.6% of the sample firms pay dividends and these firms have mean industry cash flow risk of 0.109, mean market-to-book ratio of 2.121, and mean 20

firm size of 4.577. Table 3 presents the correlation matrix, which shows the pairwise relationship between the variables. The table shows that firm cash holdings are negatively related to total institutional ownership, though the coefficient is small and statistically insignificant (p-value = 0.855). Firm cash holdings are positively related to short-term institutional ownership while negatively related to long-term institutional ownership. The coefficients are all statistically significant at the one-percent level. Firm cash holdings are also positively correlated to the institutional ownership dummy, industry cash flow risk, market-to-book, R&D and negatively related to firm size, cash flow, net working capital, capital expenditure, leverage, dividend dummy and acquisition. 4. Empirical Results 4.1. Univariate Analysis In this section, we conduct univariate analysis on the relationship between institutional ownership and firm cash holdings. More specifically, we include observations with missing institutional ownership information (i.e., zero total institutional ownership) into a portfolio called P0. The remaining observations are sorted into ten equal portfolios based on their total, short-term and long-term institutional ownership, respectively. P1 is the portfolio with the lowest level of institutional ownership, while P10 is the one with the highest level. Panel A of Table 4 presents the mean cash holdings for each portfolio. Column (1) of Panel A presents mean cash holdings for portfolios grouped by total 21

institutional ownership. The results show that observations with missing institutional ownership data (P0) have mean cash holdings of 0.160. For the remaining observations, mean cash holdings increase with total institutional ownership and reach as high as 0.183 in P5. After that, it decreases gradually to 0.150 in P9. Mean cash holdings increases again to 0.162 in P10. Therefore, no monotonic relationship is observed between total institutional ownership and firm cash holdings. Since institutional investors have varying features, different types of institutions may have varying or even opposite impact on cash holdings, resulting in the nonmonotonic relationship. Recognizing the heterogeneity among institutional investors, we sort portfolios by short-term and long-term institutional ownership, respectively. Column (2) presents the results for portfolios grouped by short-term institutional ownership. It reveals an increasing trend in portfolio cash holdings from P1 to P10, though slight fluctuations are observed midway. Column (3) presents the results for portfolios grouped by long-term institutional ownership. Again, no monotonic relationship is observed. Portfolio cash holdings increase first and then decrease from P5. But a general decreasing trend is observed for these portfolios. The results in Panel A show that portfolios grouped by short-term and long-term institutional ownership exhibit different trends in mean cash holdings. But what happens if the interaction between these two types of institutional ownership is considered? We sort observations with nonmissing institutional ownership data into five equal portfolios based on their short-term and long-term institutional ownership, respectively. P1 is the portfolio with the lowest levels of short-term/long-term 22

institutional ownership, while P5 is the one with the highest levels. Panel B of Table 4 presents the portfolio cash holdings for the 25 portfolios resulting from this two-way sorting. Numbers in parentheses are the number of observations for each portfolio. The two-way table shows that within each short-term institutional ownership portfolio, mean cash holdings decrease in general with an increase in long-term institutional ownership. For example, for the middle portfolio in Row (3), mean cash holdings start at 0.191. It increases slightly to 0.192 with an increase of long-term institutional ownership and then decreases sharply to 0.129. Further, for long-term institutional ownership portfolios, mean cash holdings increase in general with an increase in short-term institutional ownership. For example, for the middle portfolio in Column (3), mean cash holdings are 0.142 when short-term institutional ownership is low. It increases gradually and reaches as high as 0.209. Therefore, compared with the one-way sorting in Panel A, the two-way sorting displays a clearer picture of the relationship between firm cash holdings and the two types of institutional ownership. When we move from the lower-left to the upper-right of the table, a clear decreasing trend is observed. In summary, the univariate analysis documents a positive association between short-term institutional ownership and firm cash holdings, and a negative association between long-term institutional ownership and firm cash holdings. However, the findings are obtained without controlling for other determinants of firm cash holdings. These determinants may be correlated with institutional ownership and lead to spurious relationships. We leave this issue to the next section, where regression 23

analysis is conducted. 4.2. Regression Analysis In this section, we conduct regression analysis on the relationship between institutional ownership and firm cash holdings. The baseline regression specification is as follows: CASH IO IODUM INDRISK MB i, t i, t 1 i, t 2 j, t 3 i, t SIZE CF NWC CAPEX LEV 4 i, t 5 i, t 6 i, t 7 i, t 8 i, t DIVDUM RND ACQ Yr 9 i, t 10 i, t 11 i, t t i, t (5) where i indexes firm, t indexes year, j indexes industry, Yr is year fixed-effect and ε is the error term. The dependent variable is firm cash holdings (CASH) and the independent variable of interest is institutional ownership (IO). Control variables include institutional ownership dummy (IODUM), industry cash flow risk (INDRISK), market-to-book ratio (MB), firm size (SIZE), cash flow (CF), net working capital (NWC), capital expenditure (CAPEX), leverage (LEV), dividend dummy (DIVDUM), R&D (RND), and acquisition (ACQ). The regression is performed by pooled ordinary least squares (pooled-ols), with the t-statistics computed using standard errors robust to both clustering at the firm level and heteroskedasticity. The regression results are presented in Table 5. Column (1) of Table 5 shows that total institutional ownership is positively related to firm cash holdings. The coefficient is statistically significant at the one-percent 24

level. The magnitude of the coefficient indicates that a one unit change in institutional ownership leads to 0.016 unit changes in firm cash holdings. Moreover, the coefficient on the institutional ownership dummy is positive and significant as well, suggesting that firms with positive institutional ownership (i.e., nonmissing institutional ownership information) in general have more cash holdings than those with zero institutional ownership (i.e., missing institutional ownership information). At first glance, the results are consistent with both the trading hypothesis and Type 2 monitoring hypothesis. The results for the control variables are largely consistent with prior studies of firm cash holdings (e.g. Opler et al. (1999) and Bates, Kahle, and Stulz (2009)). The level of cash holdings is positively related to industry cash flow risk, market-to-book, cash flow and R&D, while negatively related to firm size, net working capital, capital expenditure, leverage, dividend dummy and acquisition. Firms in industries with high cash flow risk tend to stockpile cash for precautionary purposes, as adverse cash flow shocks may result in insufficient cash for payments and investments. Growth firms with high market-to-book ratio and R&D expenses have more investment opportunities in the future. These firms hoard more cash to reduce the risk of future financing constraints. Further, large firms hold less cash due to the economies of scale in cash holdings. Firms with high net working capital also hold less cash because assets held as net working capital are substitutes for cash. Firms with greater operating cash flow are able to accumulate more cash, while those with high capital expenditure and acquisitions spend more cash and thus have lower cash reserves. Last, 25

highly levered firms and dividend paying firms have smaller cash holdings. These firms distribute cash to investors periodically and thus have less cash remaining. Moreover, dividend paying firms are likely to be mature firms with lower future financing needs. They also have better access to capital market and thus have weaker precautionary demand for cash. To differentiate between the hypotheses, we conduct analysis on short-term and long-term institutional ownership separately rather than considering them as a whole. The results are presented in Columns (2)-(5) of Table 5. Column (2) shows that short-term institution ownership is positively related to cash holdings. The coefficient is highly significant and has much larger magnitude than that of total institutional ownership. The coefficient for long-term institutional ownership in Column (3) is significant as well, but the sign is negative. Next, we include the two types of institutional ownership in the same regression and the results in Column (4) show that signs of the coefficient for short-term and long-term institutional ownership remain unchanged. The magnitude of the coefficients is even larger than that in Columns (2) and (3). A one-percentage point increase in the percentage of shares held by short-term institutional investors causes a point increase in cash holdings by 0.09 percent. This is equivalent to a 0.55% growth in cash holdings for a typical firm with cash holdings equal to the mean value of the sample firms (16.4%). By contrast, a one-percentage point increase in long-term institutional ownership causes a decrease in cash holdings by 0.11 percent, which is equivalent to 0.67% reduction in cash holdings for the typical firm. In Column (5), we include both total and short-term 26

institutional ownership into the same regression and find that after controlling for short-term institutional ownership, total institutional ownership becomes negatively related to cash holdings. The results demonstrate that the positive relationship between total institutional ownership and cash holdings is purely driven by the ownership of short-term institutions. The classification method by Yan and Zhang (2009) defines short-term institutions as those with high portfolio turnover and thus short investment horizon for a particular stock. These institutions focus on short-term trading profits and play the major role in aggressive trading activities. On the other hand, long-term institutions are defined as having low portfolio turnover and thus long investment horizon for a particular stock. These institutions focus on long-term holding profits (e.g., dividends and capital appreciation) and provide stable capital to firms they held. Bushee and Noe (2000) document that the aggressive trading activities of high-turnover institutions magnify stock return volatility, while the stable holdings of low-turnover institutions mitigate stock return volatility. Therefore, the results presented in Table 5 are consistent with the trading hypothesis that the ownership of short-term institutions (i.e., aggressive trading institutions) increases the firm's stock market risk and makes its external financing more difficult. This induces the firm's precautionary demand for cash. In contrast, the ownership of long-term institutions stabilizes the firm's stock price and reduces its precautionary demand for cash. Furthermore, the results in Table 5 are partly consistent with Type 1 monitoring hypothesis but inconsistent with Type 2 monitoring hypothesis. Due to different 27

investment horizons, short-term and long-term institutions have differing impact on corporate governance.. Because of their short holding period, short-term institutions do not have monitoring incentives. They incur costs but benefit little from monitoring activities as the effect of these activities may only be reflected in stock prices over time. Moreover, short-term institutions do not have enough time to become familiar with the firm's operations, which further reduces their ability to take corrective actions for management misconduct. Therefore, the results for short-term institutional ownership are inconsistent with the Type 1 monitoring hypothesis. However results for long-term institutional ownership are consistent with this hypothesis as they are more likely to monitor management and reduce the free cash flow problem. These institutions hold stocks in anticipation of capital appreciation, have enough power over management, and enjoy the benefits of the improvement of operating performance. 3 The overall results are inconsistent with the prediction of the Type 2 monitoring hypothesis which suggests that the active monitoring of institutional owners inhibits self-interested managers from deploying cash quickly through value-destroying capital expenditure and acquisitions. 5. Robustness Checks 5.1. Endogeneity 5.1.1. Omitted Variables One criticism of the OLS estimates is that they may be biased and/or inconsistent in 3 Gaspar, Massa, and Matos (2005) and Chen, Harford, and Li (2007) show that institutional monitoring is mainly performed by institutions with long-investment horizon. 28

the presence of omitted variables. There may be unobservable firm characteristics that drive firm cash holdings but are not captured by the current controls. These characteristics make the error term correlated with the explanatory variables, which violates OLS assumptions (Wooldridge (2002)). To control for the unobserved firm characteristics, we adopt fixed-effect panel and random-effect panel regression, both of which are estimated by generalized least squares (GLS). In general, the fixed-effect estimator subtracts the time averages from the corresponding variables, while the random-effect estimator subtracts a fraction of that time average, where the fraction depends on the variance of the fixed-effect and error term, as well as the number of time periods. Hausman test shows that random-effect panel regression is more efficient than fixed-effect panel regression in our specific case. The results are presented in Table 6 and show that the positive relationship between short-term institutional ownership and cash holdings is robust to both random-effect and fixed-effect panel regressions. However, the results for long-term institutional ownership are not quite as robust. Columns (1)-(3) show that the coefficient of long-term institutional ownership remains negative and significant under random-effect panel regression, though the statistical significance declines in Column (2). In Columns (4)-(6) where fixed-effect panel regression is performed, long-term institutional ownership becomes insignificantly related to firm cash holdings. The insignificance can be explained by the nature of long-term institutional ownership. Since long-term institutions have low portfolio turnover, their ownership is relatively stable. Once time averages are subtracted, the variable becomes time-invariant, which 29

reduces its explanatory power in the regression. 5.1.2. Reverse Causality and Simultaneity Another possible explanation for the observed relationship between institutional ownership and firm cash holdings is that institutional investors may be attracted by firms with certain level of cash reserves. More specifically, short-term institutions may include cash-rich firms into their portfolios which leads to the positive relationship with cash holdings, while long-term institutions may prefer cash-poor firms and thus are negatively associated with cash holdings. Intuitively, short-term institutions focus on short-term trading profits and do not much care about the firm's long-term operations. Bushee (2001) finds that institutions with short investment horizon price firms myopically, overweighting short-term earnings potential and underweighting long-term earnings potential. Therefore, high cash holdings may not attract particular attention from short-term institutions. They may even prefer cash-poor firms as these firms are more risky and may generate higher short-term profits. Long-term institutions, on the other hand, care more about the firm's long-term operation and may tend to include cash-rich firms in their portfolio because these firms are less risky and have more investment funds when positive NPV projects emerge in the future. Therefore, long-term institutions are more likely to include cash-rich firms rather than cash-poor ones into their portfolio. Further, the analysis in Section 4 may also suffer from simultaneity problem, which arises when one or more of the independent variables is jointly determined with the 30

dependent variable, typically through an equilibrium mechanism (Wooldridge (2002)). To the specific case of this study, it is likely that the relationship between institutional ownership and cash holdings is driven by the correlation of the two variables with other firm characteristics (both observable and unobservable). We adopt a simultaneous equation approach with firm cash holdings and institutional ownership as the two endogenous variables to address the two problems. We estimate the simultaneous equations using two-stage least squares procedure (2SLS) as follows: CASH STIO LTIO IODUM INDRISK i, t 1 1 i, t 2 i, t 1 i, t 2 j, t MB SIZE CF NWC CAPEX LEV 3 i, t 4 i, t 5 i, t 6 i, t 7 i, t 8 i, t DIVDUM RND ACQ Yr 9 i, t 10 i, t 11 i, t t i, t (6a) STIO / LTIO CASH IODUM MB SIZE i, t i, t 2 1 i, t 1 i, t 2 i, t 3 i, t LEV PROF RET ANAFOL SHO 4 i, t 5 i, t 6 i, t 7 i, t 8 i, t TRDVOL SPREAD Yr u 9 i, t 4 i, t t i, t (6b) where PROF is the ratio of operating income before depreciation over book value of assets, RET is one-year cumulative stock return, ANAFOL is the natural logarithm of the number of earning-per-share (EPS) estimates by analysts, SHO is the natural logarithm of the number of shares outstanding, TRDVOL is the median monthly trading volume over shares outstanding, SPREAD is the difference between bid and ask stock prices over average bid/ask stock prices, and all other notation is the same as Equation (5). Stock price information comes from the CRSP Monthly database and 31