Corporate Liquidity Management and Financial Constraints

Transcription

1 Corporate Liquidity Management and Financial Constraints Zhonghua Wu Yongqiang Chu This Draft: June 2007 Abstract This paper examines the effect of financial constraints on corporate liquidity management by investigating how firms adjust cash holdings and bank line holdings in response to their cash flows. We first present a simple liquidity demand model, in which firms face costly external finance and future investment uncertainty. To hedge against potential capital shortfalls, firms design cash policy and bank debt policy to ensure financial liquidity based on their financial conditions. The model predicts that constrained firms will increase cash holdings but reduce bank line holdings when they experience positive cash flow innovations, while unconstrained firms do not exhibit such a pattern. Using simultaneous equation systems, we then test these predictions based on a unique sample of real estate investment trusts (REITs). The results strongly support our predictions. In addition, we provide evidence showing that dividend policy is sticky and plays a passive role in corporate liquidity management. We are grateful to Timothy Riddiough, François Ortalo-Magné, James Seward and James Shilling for their helpful comments. In addition, We benefited from discussion with Jim Clayton, Mike Mihelbergel, Toni Whited and seminar participants at the University of Wisconsin-Madison. Department of Finance, Florida International University, RB 208, SW 8th Street, Miami, FL 33199; Wuz@fiu.edu. Department of Real Estate and Urban Land Economics, University of Wisconsin-Madison, 975 University Avenue, Madison., WI 53706; Yongqiangchu@wisc.edu.

2 1 Introduction Corporate liquidity management has been a growing research area in corporate finance during the past ten years. Meanwhile, the effect of financial constraints on corporate behaviors remains to be a topic of continued interest. 1 Despite the extensive research on each subject, few studies have been done by combining these two research lines to examine the effect of financial constraints on corporate liquidity management. 2 Previous literature in corporate liquidity management and financial constraints have largely concentrated on the role of cash holdings (we call it as internal liquidity ). For instance, Opler et al. (1999) examine the determinants and implications of holdings of cash and find that firms with strong growth opportunities and riskier cash flows hold more cash. Almeida, Campello, and Weisbach (2004) show that the effect of financial constraints can be captured by the firm s propensity to save cash out of cash flows ( cash flow sensitivity of cash ). However, these studies do not consider the role of bank lines of credit (we call it as external liquidity ). 3 Given that bank lines of credit serve as a viable liquidity substitute to firms and help reduce capital market frictions (Holmstrom and Tirole (1998)), it would be surprising if one does not take into account the external liquidity when studying the effect of financial constraints on corporate liquidity management. A recent paper by Sufi (2006) examines the factors that determine whether firms use bank lines of credit or cash in corporate liquidity management. The author finds that firms with low cash flow are less likely to obtain a line of credit and thus rely more heavily on cash. Sufi s paper is one of the first empirical studies on the role of bank lines of credit in corporate finance, however, it does not explicitly examine how firms optimally manage financial liquidity using both cash and bank lines of credit. Rather, Sufi focuses on firms 1 A partial list of the liquidity management literature includes Kim, Mauer, and Sherman (1998), Holmstrom and Tirole (1998), Opler et al. (1999), and Faulkender and Wang (2006). Also, the representative studies in the financial constraint literature include Fazzari, Hubbard, and Petersen (1988) and Almeida, Campello, and Weisbach (2004). 2 Indeed, as pointed out by Almeida, Campello, and Weisbach (2004), corporate liquidity management and firms financial conditions are closely linked. If firms have unlimited access to external capital markets at fair prices, financial liquidity is irrelevant. In contrast, if firms are financially constrained, liquidity management becomes an important issue in their investment and financial policies. 3 In the case of Almeida, Campello, and Weisbach (2004), their focus is to develop a new test of the effect of financial constraints on corporate behaviors in general, instead of investigating the effect of financial constraints on corporate liquidity management. 1

3 cross-sectional variation in use of cash or bank lines and concludes that use of bank lines of credit is, to some extent, determined by banks based on firms cash flows. As documented in Sufi (2006), over 80% of firms across different industries have access to bank lines of credit during the period of Given this fact, it is interesting to ask how these firms manage financial liquidity by holding cash and acquiring bank lines in order to hedge against future capital shortfalls for potential investment. In particular, one can argue that, although banks have certain degree of power in controlling firms use of bank lines, essentially it is firms themselves that design optimal cash policy and bank debt policy to meet financial liquidity needs. Moreover, for the majority of firms, a key determinant in financial liquidity management is the hedging cost associated with liquid capital from different sources. Thus, one would expect that these firms should manage financial liquidity to hedge against future capital shortfalls while minimizing hedging costs based on their financial conditions. This paper investigates how firms with different financial conditions adjust two sources of liquid capital (i.e., cash holdings and bank line holdings) in response to their cash flows. Our focus is to understand the effect of financial constraints on corporate liquidity management. 4 We emphasize the hedging perspective of corporate liquidity management, that is, firms dynamically adjust cash holdings and bank line holdings to hedge against potential capital shortfalls. 5 The effect of financial constraints is captured by the different responses of the two sources of liquidity to cash flow innovations by firms with different financial conditions. Besides, we also explore issues related to dividend policy in the context of corporate liquidity management. 6 Dividend policy is relevant here because paying out more dividends reduces firms internal cash flow and thus affects their liquid capital holdings. By examining dividend 4 Specifically, we consider that a firm is financially constrained when it faces a cost gap between internal cash and external capital due to capital market frictions. That is, for financially constrained firms, the cost of obtaining external funds is significantly higher than that of internal cash. According to this classification, perhaps a large portion of firms are financially constrained. However, we believe it is difficult to exactly distinguish constrained and unconstrained firms. Instead, it is the degree of financial constraint that matters, which depends on firms financial conditions. 5 For this purpose, we measure financial liquidity using unused bank lines and total cash holdings at the end of the fiscal year. 6 A common feature shared by the previous studies is that the role of dividend payout in corporate liquidity management is often ignored. 2

4 payout behaviors in the context of liquidity management, we provide new insights about corporate dividend policy. We first present a simple liquidity demand model in which firms face capital market frictions and future investment uncertainty. Since it is costly for firms to issue securities through the public capital markets when they need capital to fund new investment projects (Myers and Majluf (1984)), firms desire to hold liquid capital ex ante to hedge against future capital shortfalls (Kim, Mauer, and Sherman (1998)). Moreover, it is assumed that firms can endogenously determine how much cash or bank lines of credit to hold to ensure financial liquidity. However, the hedging costs of holding cash and bank lines of credit vary for firms with different financial conditions. Thus, firms have to design cash policy and bank debt policy based on their financial conditions to ensure financial liquidity while controlling the hedging costs. The main prediction from the model is that financially constrained firms will reduce bank line holdings but increase cash holdings when experiencing positive cash flow innovations. That is, there is a positive relation between cash holdings and realized cash flows ( cash flow sensitivity of cash ) and a negative relation between bank line holdings and realized cash flows ( bank line sensitivity of cash ) for constrained firms. In contrast, for those unconstrained firms, there is no such a systematic pattern. We argue this stark contrast reflects the effect of financial constraints on corporate liquidity management. Moreover, our model implies that when constrained firms increase cash holdings, they do not increase dividend payout simultaneously. However, it is not necessarily the case for unconstrained firms. We then use a sample of Real Estate Investment Trusts (REITs) to test our predictions. REITs provide us with a natural laboratory to conduct the empirical tests for the following reasons. First, by tax law REITs have to pay out 90% of taxable income in form of dividend to shareholders, which limits their abilities to retain cash. 7 Consequently, they have to 7 In this sense, REITs are considered financially constrained because of the exogenous dividend restriction. However, they do have some discretionary power in retaining cash. Wang, Erickson and Gau (1993) documents that the 90% restriction is not a binding constraint for REITs, although it limits their ability to retain cash, because REITs cash flow is often much higher than the taxable income due to large depreciation write-off of real assets. Moreover, Kallberg, Liu, and Srinivasan (2003) shows that REITs pay out 60%-85% of Funds From Operations (FFOs) as dividend. 3

5 carefully manage financial liquidity by both holding cash and acquiring bank lines in order to meet their capital needs. Second, REITs are more transparent than other firms (Capozza and Seguin (1999)). Their cash flows are largely determined by current rental leases thus contain less noisy information (Ruah (2004)). Finally, REITs are public traded firms with most of the assets being tangible, and default risks of REIT bank debt are fairly low. Also, REITs generally have a higher level of before-dividend cash flows. As such, access to bank lines is not an issue for most of the REITs. All of the characteristics of REITs contribute to a unique environment for examining the roles of cash and bank lines in providing financial liquidity and the effect of financial constraints on corporate liquidity management. As firms are likely to determine their cash holdings and bank line holdings jointly, we use simultaneous equation models to capture the joint determination process of the two liquid capital holdings (Acharya, Almeida, and Campello (2006)). 8 Specifically, a two-equation system (cash holdings and bank line holdings equation) and a three-equation system (cash holdings, bank line holdings, and dividend equation) are estimated, based on both the full sample and the subsamples classified by financial constraint criteria. 9 Overall, the results strongly support our predictions. Constrained firms save cash out of cash flows ( cash flow sensitivity of cash ) and reduce bank line holdings when experiencing positive cash flow innovations ( bank line sensitivity of cash ). In contrast, no such a systematic pattern exists for those unconstrained firms. In addition, the results show that, when constrained firms increase cash holdings, they do not increase dividend payout simultaneously. This paper provides new understandings about corporate liquidity management and how financial constraints affect corporate policies. Specifically, we distinguish two sources of liquid capital (cash and bank lines of credits) and examine optimal responses of these liquid capital to cash flow innovations to understand the effect of financial constraints on corporate liquidity management. To a large extent, the estimation bias problem in traditional financial constraint literature is sidestepped (Fazzari, Hubbard, and Petersen (1988), and Hoshi, 8 Acharya, Almeida, and Campello (2006) focus on the interplay between cash policy and general debt policy, however, they do not explicitly examine bank line policy. 9 The details on these criteria will be discussed in the next section. 4

6 Kashyap, and Scharfstein (1991)) 10 and our results provide further support to Almeida, Campello, and Weisbach (2004). We argue that the bank line sensitivity of cash, jointly with cash flow sensitivity of cash, serve as good metrics to identify the effect of financial constraints on corporate policies. Moreover, dividend policy, along with cash policy and bank debt policy, is examined in the context of corporate liquidity management. Previous literature in corporate liquidity often ignore dividend policy. In this paper, the interactions among cash policy, bank debt policy, and dividend policy are explored. We present evidence showing that dividend policy is sticky and plays a passive role in corporate liquidity management. The rest of the paper is organized as follows. The next section presents and analyzes a liquidity demand model that provides directly testable empirical predictions. The empirical results and their interpretations based on the theoretical model are presented in the third section. Finally, we conclude and discuss possible directions for future research. 2 The Model In this section, we set up a simple liquidity demand model to illustrate the effect of financial constraints on firms liquidity management. The model is a representation of a dynamic problem of a firm facing investment and financing decisions in imperfect capital markets. In this setting, we analyze how a firm adjusts cash holdings and bank line holdings in response to cash flow innovation. While our model is in the spirit of Almeida, Campello, and Weisbach (2004), there exist important differences between these two models. First, their model focuses on the role of cash holdings while we consider both cash holdings and bank line holdings as viable liquidity sources. Moreover, in their model, it is the tradeoff between the marginal benefit and marginal cost of holding cash (i.e., use cash to invest now or later) that drives the central results, while in our model different hedging costs of liquid capital influence firms decision in holding cash or bank lines in the presence of financial constraints. 10 Previous literature on financial constraint has focused on relationships between investment and cash flow ( investment sensitivity of cash ). This approach has been criticized on both theoretical ground and empirical basis (Kaplan and Zingales (1997), Erickson and Whited (2000)). 5

7 2.1 Structure In our setup, a firm faces costly external finance due to the capital market imperfections, and its future investment opportunities are uncertain. Thus, the firm is concerned about maintaining a desirable level of liquid capital to hedge against potential capital shortfalls when profitable investment opportunities arise. Financial liquidity can be important to the firm because of the capital markets frictions (Kim, Sherman, and Mauer (1998)). In other words, due to informational asymmetry, a firm pays significant deadweight costs for security offerings when it has to secure capital to fund investment projects (Myers and Majluf (1984)). If the firm does not issue securities, it may miss the profitable investment opportunity and thus suffers the under-investment problem. However, holding bank lines of credit can solve this problem (Holmstrom and Tirole (1998)), as the liquid capital can be used to fund investment projects without incurring deadweight costs. As such, the firm can invest in any profitable projects to maximize its value. There are two ways that a firm can maintain liquid capital holdings. It can hold cash or acquire lines of credit from banks. 11 However, the costs associated with the two liquid capital sources are different for firms with different financial conditions. Specifically, we assume that the cost difference between holding cash and holding bank lines is α. 12 way to justify the cost difference is that the firm has to pay a commitment fee ex ante to maintain a bank credit facility for future use. One In this case, a natural question to ask is why the firm wants to hold more expensive bank lines relative to cash. The main reason is that if an investment opportunity arises, the capital required is always larger than the firm s cash holdings. Besides, the firm prefers to borrow from bank lines as these credit facilities provides quick access to capital. There are three dates, 0, 1 and 2. Assume that the firm has assets in place, which produces a deterministic cash flow c 0 at date 0 and a stochastic cash flow c 1 at date 1. Let c 1 = c 1H if the state tomorrow is H, and c 1 = c 1L if the state tomorrow is L. Assume 11 In this case, we assume that the firm have access to bank lines of credit, which is different from Sufi (2006), where a firm may not be able to access bank lines due to low cash flow. 12 Kim, Sherman, and Mauer (1998) consider that holding cash is costly because cash earns a lower return compared to project return. Here, our emphasis is the cost difference between internal liquidity and external liquidity, so we normalize the cost of holding cash to be zero and make the cost difference to be α. 6

8 c 1H > c 1L. With probability p, the state will be H, and with probability (1 p), the state will be L. At date 0, the firm has access to an investment opportunity that requires I 0 today and pays off g(i 0 ) at date 2. Additionally, the firm expects to have access to another investment opportunity at date 1, which requires capital input I 1. Assume I 1 > C. The payoff of date 1 investment I 1, which will be realized at date 2, is f(i 1 ). The firm chooses to finance investment either through internal cash or debt borrowed from bank lines of credit. The internal cash comes from its cash holdings. In order to borrow from bank lines, the firm has to maintain its credit facility by paying a total upfront commitment fee of αb where B is the upper limit of its bank lines or capacity of bank lines. Assume α << Analysis The firm s objective is to maximize the expected sum of dividends subject to various budget and financial constraints. The problem can be written as s.t. max d 0 + [pd 1H + (1 p)d 1L ] + [pd 2H + (1 p)d 2L ] (1) C,B,d,I,b d 0 = c 0 I 0 C αb (2) d 1S = c 1S + b 1S I 1S + C, for S=H, L (3) d 2S = g(i 0 ) + f(i 1S ) b 1S, for S=H, L (4) b 1S B, for S=H, L (5) where C cash holdings B total bank line capacity (i.e., unused bank lines) d 0 dividend at date 0 d 1S dividend at date 1 if the date 1 state is S d 2S dividend at date 2 if the date 1 state is S b 1S bank lines drawn at date 1 if the state is S 7

9 2.2.1 Unconstrained Firm To start the analysis, we first investigate the case where the firm is financially unconstrained. Under this circumstance, the firm can always acquire liquid capital from the capital markets to fund its investment projects. In other words, we can think of the cost difference between internal cash and external finance, α being equal to zero. Thus, whenever the firm has a profitable investment opportunity, it can obtain liquid capital from the capital markets at a fair price to fund the investment. In this case, financial liquidity is irrelevant. We can show that, for an unconstrained firm, it is able to invest at the first-best level both at time 0 and 1. Thus, we have the following first-order conditions: g (I F B 0 ) = 1 (6) f (I F B 1S ) = 1 (7) To achieve first-best investment level, the firm has to be unconstrained, i.e. it does not incur more costs to raise external funds than internal cash. Thus, the following conditions must be satisfied: c 0 > I F B 0 + C F B (8) c 1S + C F B > I F B 1S (9) A necessary but sufficient condition for firm to be unconstrained is c 0 + c 1S > I F B 0 + I F B 1S (10) Constrained Firm In the case where the firm is financially constrained, it has to borrow from bank lines to finance its investment projects, due to insufficient cash holdings and inability to secure other source capital from the capital markets at fair prices. Under these circumstances, we can 8

10 show that the firm will not pay dividends at date 0 and date 1, i.e., d 0 = 0, and d 1S = Another observation is that b 1L = B, given that the firm s cash flow is low in L state, it tends to use up the full capacity of its bank lines. Under these assumptions, the firm s problem can be rewritten as: max C,B,b 1H g(i 0 ) + p[f(i 1H ) b 1H ] + (1 p)[f(i 1L ) B] (11) s.t. I 0 = c 0 C αb (12) I 1L = c 1L + C + B (13) I 1H = c 1H + C + b 1H (14) The solution can be characterized by the following first-order conditions: g (I 0 ) = pf (I 1H ) + (1 p)f (I 1L ) (15) g (I 0 )α = (1 p)(f (I 1L ) 1) (16) f (I 1H ) = 1 (17) Some observations from the above first-order conditions are: (1) In state H, the firm can achieve first best investment, which follows from equation (17); (2) I 1L < I F B 1L, which follows from equation (16) that f (I 1L ) > 1; (3) I 0 < I F B 0, which follows from (15) that g (I 0 ) = p + (1 p)f (I 1L ) > p + (1 p) = 1. We now state and prove the central results of our model. That is, Proposition 1: For firms that are financially constrained, their liquidity management policy depends on the realized cash flows in the following way. (1) The optimal cash holdings, C, increase with realized date 0 cash flows c 0, i.e., C/ c 0 > 0, the positive cash flow sensitivity of cash; 13 Alternatively, we can normalize the firm s dividend payout and assume that the firm will maintain the same level of dividend it pays during the previous period. In this case, we have d 0 = d 0 d = 0, d 1S = d 1S d = 0. This will not change the final results of our model. 9

11 (2) The optimal bank line holdings, B, decrease with date 0 cash flow c 0, i.e., B/ c 0 < 0, the negative bank line sensitivity of cash. Proof: Combine (15) and (16), we get, (1 α)g (I 0 ) = 1 (18) Differentiate both sides of (18) with respect to date 0 cash flow c 0, we get, (1 α)g (I 0 )(1 C c 0 α B c 0 ) = 0 (19) thus, C c 0 + α B c 0 = 1 (20) Then differentiate both sides of (16), we get, (1 p)f (I 1L )( C c 0 + B c 0 ) = 0 (21) thus, C + B = 0 (22) c 0 c 0 Combine (20) and (22), we get, C = 1 c 0 1 α > 0, since α < 1 (23) and B = 1 c 0 1 α < 0 (24) The basic intuition from the above results can be summarized as follows. For constrained firms, they have to carefully manage cash holdings and bank line holdings to ensure financial liquidity while controlling the hedging costs. Because external finance is more expensive to them, reducing financing costs is an important factor in their liquidity management. Specifically, as the hedging costs of using bank lines are higher than those of using cash, these firms would rationally save more cash out of cash flow, instead of holding more bank lines, when experiencing positive cash flow innovations. However, for unconstrained firms, external 10

12 liquidity is less expensive and they are more concerned about hedging against potential capital shortfalls. Hence, liquid capital holdings of unconstrained firms are expected to be less responsive to cash flow shocks. Based on Proposition 1, the first hypothesis is stated as follow. Hypothesis 1: Financially constrained firms will increase cash holdings but reduce bank line holdings when experiencing positive cash flow innovations, i.e., there exists positive cash flow sensitivity of cash and negative bank line sensitivity of cash. However, unconstrained firms will not exhibit such a pattern. Specifically, for constrained firms, we have, ( C c 0 )constrained > 0, ( B c 0 )constrained < 0 (25) In addition, the analysis above implies that constrained firms will restrain their dividend payout when they anticipate new investment opportunities, since paying more dividend reduces the firm s financial liquidity and add to the need for external liquidity. Similarly, when constrained firms increase their cash holdings, they do not acquire more bank line holdings simultaneously. Thus, we expect that constrained firms would not increase dividend payout when they increase cash holdings. The second hypothesis is stated as follows. Hypothesis 2: Constrained firms do not increase dividend payout and bank line holdings when they increase their cash holdings. However, for unconstrained firms, it may not be the case. 3 Empirical Tests In this section, empirical tests are conducted based on the hypotheses from the previous section. The focus is on how firms cash holdings and bank line holdings respond to the fluctuation of their cash flows. Both the full sample and the subsamples of REITs classified by various financial constraint criteria are used to show the effect of financial constraints on corporate liquidity management. In addition, the role of dividend payout in the corporate liquidity management is also examined. 11

13 3.1 Methodology We follow the literature of cash flow sensitivity of cash (Almeida, Campello, and Weisbach (2004)) to specify our estimation equations. In their paper, a general form of cash equation is estimated as follows: C it = β 0 + β 1 CF it + β 2 T Q it 1 + ɛ it (26) where the dependent variable is the change of C it from year t-1 to year t ( C it ), and C it is the cash holdings scaled by total assets. The independent variables include CF it (the cash flow measure), scaled by total assets, and T Q (the Tobin s q measure). When different samples of firms based on certain financial constraint criteria are estimated, β 1 can be used as an indicator of the effect of financial constraints on corporate policies. As cash holdings and bank line holdings are likely to be determined jointly by firms, a two-equation simultaneous system is used to estimate the response of liquid capital holdings to cash flow. To capture the dynamic relationships, we use change of cash holdings and change of bank line holdings as the dependent variables. The key independent variable is firms cash flow. Besides, we add two variables, i.e., the bank line holdings and the cash holdings in year t-1, to identify the two equation system. A Tobin s Q measure and firm size are also included in the system as control variables. Finally, we add year and property-type fixed effects to capture the sources of variation from different years and property types of firms. As such, our primary empirical model is given as follows: L it = β 0 + β 1 NCF it + β 2 T Q it 1 + β 3 CS it + β 4 L it 1 + β 5 Inta it + ɛ it (27) CS it = γ 0 + γ 1 NCF it + γ 2 T Q it 1 + γ 3 L it + γ 4 CS it 1 + γ 5 Inta it + ξ it (28) where L it is the ratio of bank line holdings over total assets in year t, and L it is the net increase of the ratio from year t-1 to t. Similarly, CS it is the net increase of the ratio (cash holdings over total assets) from year t-1 to year t, NCF it is the total cash available after 12

14 paying out dividend in year t, scaled by the total assets. 14 D it 1 is the total cash dividend paid at the end of year t-1, scaled by total assets. Beginning-of-year Tobin s Q (T Q it 1 ), defined as the total market cap of a REIT plus total debt divided by the total assets at the end of year t-1, serves as a proxy for growth opportunity of a REIT. The β 1 and γ 1 in the equation are the coefficients of particular interest, as they indicate how a REIT adjusts bank line holdings and cash holdings based on its realized cash flow. Moreover, to examine the role of dividend policy in corporate liquidity management, we allow dividend payout to be an endogenous variable. This also serves as a robustness check for our primary two-equation estimation. Thus, we estimate a three-equation system by adding the dividend equation with the change of dividend payout as the dependent variable. That is, L it = β 0 + β 1 CF it + β 2 T Q it 1 + β 3 CS it + β 4 D it + β 5 L it 1 + β 6 Inta it + ɛ it (29) D it = β 0 + γ 1 CF it + γ 2 T Q it 1 + γ 3 CS it + γ 4 L it + γ 5 L it 1 + γ 6 Inta it + ϑ it (30) CS it = γ 0 + α 1 CF it + α 2 T Q it 1 + α 3 L it + α 4 D it + α 5 CSit 1 + α 6 Inta it + ξ it (31) where CF it is the total cash available in year t, scaled by the total assets. The bank line holdings, the cash holdings, and the dividend payout variable in year t-1, all scaled by total assets, are used to identify the three-equation system. First, the two equation and the three equation system are estimated using the full REIT sample. By so doing, we can test how the two liquid capital holdings as well as dividend payout respond to cash flow innovations. Next, we classify the full sample based on three financial constraint criteria and estimate the two systems using each of the six subsamples. The idea is to further examine the effect of financial constraints on firms financial policies, in particular, corporate liquidity management and dividend policy. 14 Essentially, NCF it is the net cash flow available. In addition, Sufi (2006) argues that, instead of using total assets as a scalar, one should use total assets minus cash holdings. Otherwise, there will a mechanical bias on the relevant coefficients. We use the alternative measure according to Sufi (2006), the estimation results are very similar to the ones otherwise. Hence, we do not report the alternative results. 13

15 3.2 Financial Constraint Criteria To classify the full sample, we follow the methodology used in the financial constraint literature (e.g., Fazzari, Hubbard, and Petersen (1988)) by splitting the full sample based on various priori measures of financial constraints. Specifically, three financial constraint criteria are used. Scheme # 1: Firm size. Firm size is considered an important criterion for financial constraint as small firms are more likely to face borrowing constraint from the public capital markets than larger firms (See Whited (1992), Almeida, Campello, Weisbach (2004)). Also, larger firms have lower external financing costs because of the scale of economies from lower fixed cost of security issuance. Thus, we sort the sample based on total assets of REITs, and assign the top one third and the bottom one third as the constrained and the unconstrained group, respectively. This classification results in 476 firm-year observations in each subsample. Scheme # 2: Bond Ratings. If a firm issues a public debt offering and has a bond rating for its public debt, it shows that the market recognizes the firm s credit quality. Thus, the firm is likely to have better access to the capital markets than those otherwise. Previous studies such as Whited (1992) often use this criterion to characterize the degree of financial constraint. Here, we classify the sample based on whether a firm issues a public debt or not. The firms that issued a public debt and obtained a bond rating are classified as the unconstrained group. There are 809 (596) firm-year observations in the constrained (unconstrained) subsample. Scheme # 3: Banking relationships. Diamond (1989) shows that firms establish relationships with banks before they can access the public capital markets. In fact, REITs rely more on bank lines of credit to obtain financial liquidity to fund investment due to their limited abilities to retain earnings. Hence, a REIT with good banking relationship is less likely to be financially constrained. Sufi (2006) argues that lack of access to a bank line is a powerful measure of financial constraints. However, we believe whether a firm can repeatedly access bank lines should be a better measure of financial constraints. If a firm develops a close banking relationship through repeated interactions with banks, the firm is likely to 14

16 have ready access to the capital markets. We follow the literature (see Bharath et al. (2005)) and classify the full sample into two groups based on a dummy variable, called banking relationship dummy. The REITs that have not established a close banking relationship are assigned into the constrained group, and those with a banking relationship are assigned into the unconstrained group. There are 750 (679) firm-year observations in the unconstrained (constrained) group. One should not be surprised that the groups based on the three different schemes 15 are to some extent correlated. For example, large firms are more likely to have a bond rating. However, based on the number of observations in each subsample, these cross correlations are not very strong. 3.3 Data and Descriptive Analysis We construct a unique data set of REITs from three different data sources: 1. SNL REIT Financial database; 2. NAREIT Capital Offering database; and 3. LPC s DealScan Commercial Loan database. The primary data used is from the SNL REIT database, which provides detailed firm classification and financial information of REITs. For example, we can identify whether a REIT is an equity REIT, and which property type a REIT focuses on. More important, besides financial information such as total assets, asset growth rate and market capitalization, the database also provides bank line usage information, i.e., bank line holdings at the end of a given year, the amount of debt drawn from bank credit lines during a given year, and average drawn ratio of bank lines for a REIT each year. To be included in our sample, a REIT has to meet the following criteria: (1) listed on NYSE, AMEX or NASDAQ, and elected REIT tax status; (2) registered with the National Real Estate Investment Trust Association (NAREIT); and (3) must be an equity REIT. The original sample from the SNL REIT database has 3,667 firm-year observations. To add REIT capital offering information, we obtain REIT capital offering data from the NAREIT database, which consists of 1,401 seasoned equity offerings, 950 public debt offerings, and 156 IPOs of REITs. Then, we hand match the capital offering data into the SNL 15 We do not choose dividend payout ratio as one criteria here as REITs have the 90% dividend payout rule which complicates REITs dividend payout behaviors. 15

17 REIT sample. The final data set consists of 1,429 REIT firm-year observations from 1990 to To classify the full sample based on banking relationship and bond rating, we obtain the bank loan information from the DealScan loan database of Loan Pricing Corporation. There are 1,248 REIT bank loans in the sample. Also, REITs bond rating information is obtained from the NAREIT s Capital Offering database. [Insert Table 1 and 2 here] Summary statistics for the full sample and the subsamples are presented in Table 1 and Table 2, respectively. There are six subsamples based on the three financial constraint criteria: firm size, bond rating, and banking relationship status. First, note that cash holdings statistics of REITs are much less than those non-reit public firms. For example, the average cash holdings of REITs is 2.19%. However, the regular public firms on average hold about 10% of cash (see Achaya et al. (2006)). Second, the constrained REITs tend to hold more cash flow compared to the unconstrained REITs. This is consistent with the notion that the constrained firms are more likely to save more cash to ensure financial liquidity for investment while the unconstrained firms may use more bank lines as a liquid capital source. Another interesting fact is that the unconstrained REITs have higher Tobin s Q than the constrained REITs. Specifically, REITs with banking relationships and large REITs have higher Tobin s Q than those without banking relationships and with small size. This finding is consistent with the argument of Han (2004). Moreover, the constrained REITs do not cut their dividend over years, i.e., D it are all positive while in general being a small percentage. This is consistent with Capozza and Seguin (1998) that REITs maintain a stable, growing dividend stream over years as cutting dividends may result in severe punishment from the stock markets. 3.4 Cash, Bank Lines, and Cash Flows - Full Sample Table 3 shows the regression results from the two simultaneous equation models based on the full sample. The first model includes two equations: bank line holdings equation and 16 Following the investment-cash flow literature (see Cleary (1999)), we winsorize the data, e.g., Tobins q measure is limited between 0 and 4. 16

18 cash holdings equation (see Column (1)-(2)). In particular, dividend payout is assumed exogenous and does not play an active role in firms liquidity management. Hence, we subtract dividend from firms cash flow and obtain the net cash flow measure (NCF ). The second model includes three equations: bank line holdings equation, cash holdings equation, and dividend payout equation (see Column (3)-(5)). Here, dividend payout is endogenized in the sense that we assume it plays an active role in firms financial decision making. By so doing, we can examine how dividend policy interacts with cash policy and bank line policy and how it is influenced by firms cash flow. [Insert Table 3 here] The results of the first model (Column (1) - (2)) show that there exists a positive relationship between firms net cash flow and the change in cash holdings. That is, REITs increase cash holdings when they receive more net cash flows. This is consistent with Almeida et al. (2004) that, as a group of constrained firms, REITs tend to save more cash out of cash flow. On the other hand, a negative relationship is observed between firms net cash flow and bank line holdings. That is, when REITs net cash flows increase, they tend to reduce bank line holdings, or equivalently, boost their debt capacities for external liquidity. The latter result is interesting and new to the literature, implying that, for constrained firms, saving debt capacities is an equivalent mechanism to ensure financial liquidity compared with saving more cash out of cash flow. Next, Column (3) - (5) of Table 3 show the results for the three equation model. We find a similar pattern regarding how bank line holdings and cash holdings respond to cash flow innovations as in the first model. That is, when REITs receive more cash flow, they tend to save more cash while reducing bank line holdings or increasing bank line capacity. These results confirm our previous findings, i.e., no matter whether we treat dividend payout endogenous or exogenous, the main results regarding liquid capital holdings and cash flow innovations are similar. In addition, Column (3) shows that the coefficient of CS is significantly negative, suggesting that when REITs increase cash holdings, they often reduce their bank line holdings. This supports the second hypothesis regarding the relationship between cash holdings and bank line holdings. 17

19 3.5 Cash, Bank Lines, and Cash Flow - SubSamples To further examine the effect of financial constraint on firms financial policies, we compare the coefficients of the key variables (cash holdings and bank line holdings) for the constrained and the unconstrained firms. First, we look at the bank line holdings equation of the two equation model. For the constrained firms across the three different groups, the coefficients of net cash flow (NCF ) are all negative and at the 1% significance level (see Panel A of Table 4), indicating that the constrained firms reduce bank line holdings upon a higher level of cash flow. However, the coefficients for the unconstrained firms are all positive and insignificant. Moreover, the coefficients of CS it are negative and significant for the constrained groups, but not for the unconstrained groups. Taken together, these sharp contrasts indicate that the constrained firms dynamically adjust their bank line holdings based on their cash flow and cash holdings, but the unconstrained REITs do not. We argue that it is the effect of financial constraints that results in such a systematic pattern on firms liquidity management behaviors. [Insert Table 4 here] Second, we examine the cash holdings equation of the two equation model. Unlike the bank line holdings equation, for the constrained firms across the three different groups, the coefficients of net cash flow (NCF ) are significantly positive (see Panel B of Table 4), suggesting that when REITs have positive cash flow innovations, they increase cash holdings. In contrast, the coefficients for the unconstrained firms are all negative. Again, these results imply that the constrained firms are more responsive to cash flow shocks than the unconstrained firms, which is consistent with Almeida, Campello, and Weisbach (2004). However, the change of bank line holdings does not seem to have a significant impact on the cash holdings. One possible explanation is that there exists an asymmetric relationship between cash holdings and bank line holdings - as one of the external finance vehicles, bank lines are more costly than cash. Consequently, constrained firms would rather hold more cash when they face investment uncertainty. Next, we examine the three equation system, in which dividend payout is considered one of the endogenous variables. In general, we find similar patterns in the relationships among 18

20 bank line holdings, cash holdings and cash flow. That is, when firms have positive cash flow innovations, the constrained firms reduce bank line holdings (see Panel A of Table 5), but increase their cash holdings (see Panel C of Table 5). In contrast, the unconstrained firms reduce cash holdings and increase bank line holdings upon receiving more cash flows. To sum up, we argue that firms financial status has a significant impact on their liquidity management, and the effect of financial constraints is captured by the systematically different liquidity management behaviors between the constrained and the unconstrained firms. [Insert Table 5 here] The other control variables in the three equations also have expected signs while some of them are not statistically significant. For example, in the bank line holdings equation, the coefficients of T Q it 1 for the constrained group are all positive and significant, suggesting that firms with good investment opportunities are more likely to increase their bank line holdings to fund investment. On the other hand, in the cash holdings equation, the coefficients of T Q it 1 for the less constrained group are all positive and significant, implying that the lessconstrained firms are more likely to increase cash holdings to ensure liquid capital when they have good investment opportunities. 3.6 The Role of Dividend Policy In this section, the role of dividend policy in corporate liquidity management is examined. First, we investigate how firms dividend payout responds to their cash flow innovation and whether or not the responses are different for the constrained and the unconstrained group. Column 4 of Table 3 shows that there is a positive relationship between cash flow and dividend payout. That is, when cash flows increase by 1%, dividend paid tend to increase by 0.30%. Moreover, the subsample estimation also confirms the same results. Panel B of Table 5 shows that the coefficients of CF it are all significantly positive, suggesting that positive cash flow innovation has a positive impact on dividend payout. This is consistent with the agency theory of dividend policy (see Wang et al. (1993)). Moreover, the interactions among dividends, cash holdings, and bank line holdings are investigated in the context of corporate liquidity management. Column 4 of Table 4 indicates 19

21 that an increase in cash holdings as well as an increase in bank line holdings are inversely related to firms dividend payout level. Specifically, when bank line holdings and cash holdings increase by 1%, their dividend payout decreases by 0.04% and 0.08%, respectively. However, dividend payout does not have a significant impact on cash holdings and bank line holdings. Specifically, while the signs of D are negative ( and -0.06, respectively), the coefficients are not significant. The results based on the subsamples are similar: Panel A and Panel C of Table 5 indicate that the coefficients of CS in the cash holdings equation and bank line equation are insignificant for the constrained firms and the unconstrained firms, while the coefficients of CS and L in the dividend equations are generally negative and significant. One possible explanation for such relationships is that, to a large extent, REITs are exogenously capital constrained. Because liquid capital is crucial for them to take quick action in property acquisition, they have to carefully manage cash holdings and bank line holdings when anticipating investment opportunities. Thus, they have less incentives to increase dividend payout during the period when they increase cash holdings and bank line holdings. Interestingly, the coefficients of CS and L are often larger for the constrained firms than for the unconstrained firms. This pattern further supports the notion that the constrained firms are more likely to limit their dividend payout than unconstrained firms when they expect good investment opportunities. These results are also consistent with the findings in Brav et al. (2005), which argue that firms prefer to set conservative dividend policy. That is, when there are more cash flows available, firms tend to increase dividend payout. However, as paying more dividend reduces firms financial liquidity available, constrained firms do not increase dividend at the cost of reducing cash flow holdings or bank line holdings. Put it another way, constrained firms do not treat dividend payout as a priority over firms liquid capital holdings and thus investment funding needs. Taken all together, these results suggest that there exist dynamic relationships among dividend policy, bank debt policy, and cash holding policy. However, dividend payout is sticky and a secondary decision relative to cash policy and bank debt policy. 20

22 In addition, we find that the change of dividend payout from year t to t-1 is negatively related to the dividend payout in year t-1. Column (4) shows that the coefficient of D it 1 is negative at the 1% significance level. That says, if firms pay relatively high dividends for the previous period, they will not significantly increase dividend payout during the next period. This is consistent with the notion that REITs might target their dividend payout. If a firm s dividend payout reaches a certain ratio, it might restrain its dividend payout level during the next period so that the firm can maintain a stable dividend stream over years. In other words, the significantly negative coefficients of D it 1 show that dividend payout of REITs in year t are highly correlated to their dividends paid in year t-1. These findings provide further support for our previous results that dividend payout is sticky and, unlike cash holdings and bank line holdings, it does not play an active role in corporate liquidity management. The other control variables in the dividend equations (see Column (4) and Panel B of Table 5) also have expected signs while some of them are not statistically significant. For example, the coefficients of T Q it 1 for the constrained group are generally positive and significant, suggesting that firms with good investment opportunities tend to increase their dividend payout. On the other hand, the coefficients of Lnta it for both the constrained group and the unconstrained group are negative, implying that larger firms are less likely to make more dividend payout. Overall, these results provide strong support for the empirical predictions regarding dividend policy. 4 Conclusion Remarks This paper examines the effect of financial constraints on corporate liquidity management by investigating how firms manage cash holdings (internal liquidity) and bank line holdings (external liquidity) based on their financial conditions to ensure financial liquidity. We present a simple liquidity demand model in which firms face costly external finance and future investment uncertainty. In this setting, firms design cash policy and bank debt policy to hedge against potential capital shortfalls for future investment projects. The model predicts that constrained firms will increase cash holdings but reduce bank line holdings when they 21

23 experience positive cash flow innovations, while unconstrained firms do not exhibit such a pattern. In addition, when constrained firms increase cash holdings and bank line holdings, they do not increase dividend payout simultaneously. We then test these predications using a unique data set from REITs. REITs provide us with a unique environment to examine these issues because they are exogenously constrained but have ready access to bank lines of credit. Based on simultaneous equation estimation method, the results strongly support our model predications, suggesting that constrained firms dynamically adjust cash holdings and bank line holdings to ensure financial liquidity based on their financial conditions, and financial constraints have a significant impact on firms cash policies and bank debt policies. Moreover, the results show that when the constrained firms increase cash holdings and bank line holdings, they do not pay more cash dividends. However, dividend payout does not have a significant impact on either cash holdings or bank line holdings. These results imply that dividend policy is sticky and plays a passive role in corporate liquidity management. This paper provides new understandings about corporate liquidity management and how financial constraints affect corporate policies. Specifically, we distinguish two sources of financial liquidity (cash versus bank lines of credits) and examine responses of these two liquid capital holdings to cash flow innovations to detect the effect of financial constraints on corporate liquidity management. To a large extent, we sidestep the estimation bias problem in traditional financial constraint literature (Fazzari, Hubbard, and Petersen (1988)) and provide further evidence to support Almeida, Campello, and Weisbach (2004). We argue that the bank line sensitivity of cash, jointly with cash flow sensitivity of cash, serve as good metrics to identify the effect of financial constraints on corporate policies. Moreover, dividend policy, along with cash policy and bank debt policy, is examined in the context of corporate liquidity management. Previous literature in corporate liquidity management often ignores dividend policy. In this paper, the interactions among cash policy, bank debt policy, and dividend policy are explored. We provide evidence suggesting that dividend policy is sticky and plays a passive role in corporate liquidity management. Admittedly, our findings and empirical results are based on a single industry - REITs, which are operated in a constrained environment. To better understand firms optimal 22