- Email: [email protected]
The link between bank monitoring and corporate dividend policy: The case of dividend omissions
Journal of Banking & Finance 25 (2001) 2069±2087 www.elsevier.com/locate/econbase
The link between bank monitoring and corporate dividend policy: The case of dividend omissions Soo-Wah Low a, Louis Glorfeld b, Douglas Hearth b, James N. Rimbey b,* b
a National University of Malaysia, Kuala Lumpur, Malaysia Department of Finance, BA-302, Sam M. Walton College of Business Administration, University of Arkansas, Fayetteville, AR 72701, USA
Received 23 November 1998; accepted 26 October 2000
Abstract This study investigates whether bank monitoring in¯uences investor response to a borrowing ®rm's decision to omit its dividend payments. We establish a new link between the theories of banking and dividend policy in an examination of how bank monitoring and ®rm dividend signals complement one another to resolve information asymmetries. Results indicate that, for small ®rms, investors interpret the dividend decision as a function of bank monitoring and the dividend signals taken together. Also reported are the results of tests examining the dierences between the monitoring eects of banks versus public and private non-bank lenders. Ó 2001 Elsevier Science B.V. All rights reserved. JEL classi®cation: G14; G21; G35 Keywords: Bank monitoring; Dividend signaling; Dividend omission; Firm size
*
Corresponding author. Tel.: +501-575-4505; fax: +501-575-8407. E-mail address: [email protected] (J.N. Rimbey).
0378-4266/01/$ - see front matter Ó 2001 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 4 2 6 6 ( 0 0 ) 0 0 1 7 0 - 9
2070
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
1. Introduction The purpose of this study is to investigate whether bank monitoring in¯uences the market's response to a ®rm's decision to omit its cash dividend. In particular, this study establishes a new link between the theories of banking and dividend policy by examining how bank monitoring and dividends as signals may complement each other in mitigating information asymmetries. Information-based theories of ®nancial intermediation suggest that banks provide a unique information production and monitoring service in an informationally asymmetric capital market. 1 These theories imply that an imperfect capital market recognizes the value of bank monitoring and uses bank debt as a signal of ®rm value. A number of empirical studies show that the presence of a bank lending relationship disseminates information about the ®rm that in¯uences the market's assessment of major corporate decisions. 2 For example, studies have shown that the presence of bank debt aects the market assessment of corporate sello decisions, the issuance of commercial paper, and seasoned equity oerings. Other studies have demonstrated a relationship between bank debt, ®rm value, and the underpricing of initial public oerings. In short, the theoretical basis and empirical evidence suggest that bank debt serves as an important signal of ®rm quality and thus contributes to reducing information asymmetries. Research also suggests that the bene®ts of banks as information specialists and monitors of corporate activity may have the greatest eect on small ®rms (e.g., Fama, 1985; Slovin et al., 1992). Whereas much information may be available about large ®rms, providing them easier access to capital markets, small ®rms are less well known. Hence, their ability to obtain credit from banks may serve as a signal of value. It is in this context that we study the interaction between ®rm size and banking relationships. In previous studies of payout policy, various theories have been set forth to explain the relationship between dividends and ®rm value. Signaling theory suggests that in a market characterized by asymmetric information, dividends can be used as a signaling device to communicate private information. In general, dividend-signaling models posit that dividend announcements convey information about the ®rm's current and/or future cash ¯ows, therefore changes in the value of the ®rm around the time of dividend announcements should be proportionate to the unexpected change in dividend policy (e.g., Bhattacharya, 1979; Miller and Rock, 1985; John and Williams, 1985).
1
Much of this literature is reviewed in Bhattacharya and Thakor (1993). See, for example, Hirschey et al. (1990), James (1987), James and Wier (1990), Slovin et al. (1988), Slovin et al. (1990) and Slovin and Young (1990). 2
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2071
While a large body of empirical evidence documents a positive association between the announcement of dividend changes and ®rm value, more recent empirical evidence calls into question the value of dividends as signals. 3 This leads to the conjecture that the market reaction to a ®rm's dividend decision may be in¯uenced by other factors, such as the presence of banks as monitors of ®rm activity. To investigate the relationship between the market reaction to dividend omission announcements and bank debt, a sample of ®rms that omitted their dividends between 1978 and 1996 is identi®ed. Daily stock price returns for each ®rm in the sample are collected along with various measures of bank debt. The sample is further divided into subsamples consisting of large and small ®rms using the same classi®cation scheme as Slovin et al. (1992). A sequence of multivariate statistical tests is then conducted to examine the role of banks as lender/monitors. Our ®ndings show a strong positive relationship between bank lending and the market's assessment of dividend omission announcements for small ®rms. Further, the market's reaction to dividend omissions by small ®rms having high levels of bank debt is far less negative than it is when small ®rms have little or no bank debt. On the other hand, for large ®rms there appears to be little relationship between bank debt and the market response to dividend omission. Our ®ndings also suggest that bank debt has a greater mitigating impact on the market response to dividend omission than either non-bank private debt or public debt. The balance of this paper is organized as follows. Section 2 discusses the relationship between bank monitoring and dividend omission announcements. Section 3 describes the research design employed by this study. Section 4 presents our empirical ®ndings and Section 5 oers our conclusions. 2. Bank monitoring and dividend omissions 2.1. Information-based theory of ®nancial intermediation Leland and Pyle (1977) develop a signaling model that emphasizes asymmetric information as a primary reason to explain the existence of ®nancial intermediaries. They argue that ®nancial intermediaries can signal the
3 Examples of studies ®nding a positive association between dividend change announcements and ®rm value include Christie (1994), Healy and Palepu (1988) and Aharony and Swary (1980). Some of the recent studies questioning the value of dividends as a signaling mechanism include Benartzi et al. (1997), DeAngelo et al. (1996), Jensen and Johnson (1995) and Christie (1994).
2072
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
proprietary information of many borrowers at a cost that is lower than that of each individual borrower attempting to signal alone. In his model of delegated monitoring, Diamond (1984) extends Leland and Pyle by showing that private information about borrowers can be monitored more eciently by a ®nancial intermediary on behalf of its depositors (as suppliers of funds) rather than by many individual depositors acting as individual lenders. Diamond provides further insight by showing that ®nancial intermediaries possess a cost advantage in producing information, and that the monitoring function provided by intermediaries is cost eective in resolving the ®nancial contracting problems caused by asymmetric information. Rajan (1992) explores the costs and bene®ts of borrowing from informed versus arm's-length sources, and oers an explanation of why ®rms may seek alternatives to bank monitoring (and control) of ®rm investment decisions. Other models of ®nancial intermediaries as delegated monitors have been developed by Ramakrishnan and Thakor (1984), Boyd and Prescott (1986), and Williamson (1986). In summary, these theoretical models of ®nancial intermediation suggest that in a market with imperfect information, ®nancial intermediaries have a comparative advantage in collecting private information and monitoring the activities of borrowers relative to the direct monitoring by many individual lenders. Fama (1985) expands this notion by distinguishing banks from other types of ®nancial intermediaries and argues that banks are indeed a special type of ®nancial institution. He draws a distinction between inside and outside debt, and classi®es bank debt as inside debt, de®ned as a contract in which the creditor (the bank) gets access to the private information about the borrowing ®rm that is not available to other market participants. This con®dential disclosure of information by managers allows banks to know more about the quality of the ®rm than any other investor. As noted by Wansley et al. (1993), and Bhattacharya and Thakor (1993), bank debt provides a stronger signal of the ®rm's value than other forms of debt because bank lenders are better informed relative to other creditors. The bank's willingness to risk its ®nancial capital by granting loans to the ®rm re¯ects positive private information about the ®rm. There is a substantial amount of empirical evidence supporting the unique nature of bank loans. James (1987), for example, ®nds that borrowers experience a signi®cant positive abnormal return upon announcing new bank loan arrangements, versus a non-positive insigni®cant abnormal return for publicly placed debt and a signi®cant negative stock price response for debt placed privately with insurance companies. Billett et al. (1995) also ®nd that bank loan announcements elicit signi®cant positive abnormal returns to the announcing ®rms. They report signi®cant positive abnormal returns for both new and renewed loans, with renewed loans generating slightly larger stock price reactions, on average.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2073
2.2. The role of banks in monitoring managerial decisions Banks may also serve another monitoring function. Studies of various managerial actions have concluded that the existence and level of bank debt in¯uences market perceptions. In general these studies show that the market uses the banking relationship as a mechanism to certify the ®rm's actions. For example, Hirschey et al. (1990) ®nd signi®cant positive returns around corporate sello announcements by companies with high bank debt. By contrast, they ®nd little market reaction to sello announcements by companies with little or no bank debt. Slovin et al. (1990) ®nd that ®rms with extensive bank ®nancing experience virtually no stock price response to announcements of seasoned equity oerings. Firms with little bank debt experience signi®cant and negative stock price responses to seasoned equity oering announcements. The presence of banks as monitors described in these studies also ®ts well with the tenets of agency theory. In each of these cases the actions of managers generate free cash ¯ow, which could be invested sub-optimally (Jensen, 1986). In eect, market participants may be reassured by the bank's presence as a third-party monitor. 2.3. Bank monitoring and ®rm size Prior studies have suggested that the monitoring and evaluation functions of banks are more important for small companies than for larger ones. Fama (1985) argues that small ®rms incur lower contracting costs by securing ®nancing from banks as opposed to obtaining ®nancing from the public debt market, which entails a high cost of disclosure. He suggests that small ®rms ®nd it cheaper to borrow from banks since they are required to supply information to fewer lenders (the banks). Large ®rms, which have less of a ®nancial contracting problem, ®nd it more economical to distribute information more widely and obtain ®nancing from the public debt market. In his model of choice between bank loans and directly placed debt, Diamond (1991) argues that small companies with credit ratings that lie toward the middle of the spectrum rely heavily on bank loans since the quality of these ®rms cannot be easily veri®ed by outsiders. As such, these ®rms have an incentive to borrow from banks in order to obtain monitoring services and to acquire reputation through time by building up their credit histories. Once that reputation is established, these ®rms can then move into the public marketplace. Empirical evidence strongly suggests that the bene®ts of banks as information specialists and monitors of corporate activity accrue primarily to small ®rms. Slovin et al. (1992), for example, ®nd that the market's reaction to the renewal or initiation of bank loan arrangements is far more pronounced in small ®rms than it is in large ®rms.
2074
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2.4. Dividend omissions and signaling The notion of the ``informational content'' of dividends was modeled formally by Bhattacharya (1979), John and Williams (1985) and Miller and Rock (1985). The common theme of these models is that in the presence of asymmetric information between managers and outside investors, dividends can be used as a signaling mechanism by managers to communicate their assessment of the ®rm's current performance and future prospects. In general, dividendsignaling models posit that dividend announcements transmit information about the ®rm's future and/or current cash ¯ows and consequently the changes in the value of the ®rm around dividend announcements should be proportionate to the changes in dividend policy. Thus, when a ®rm unexpectedly increases (decreases) dividends, it signals management's optimistic (pessimistic) outlook for the future, and based on this information, outside investors should revise their expectations about the ®rm's value. There is a substantial amount of empirical evidence that documents a positive association between the announcement of dividend changes and stock price reaction. 4 There is general agreement among these studies that the market responds positively to favorable dividend changes (initiations, increases, and resumptions) and negatively to unfavorable dividend changes (reductions and omissions). However, the available empirical evidence also indicates that unfavorable dividend changes elicit market reactions that are much greater in magnitude than favorable dividend changes. For example, Healy and Palepu (1988) show that dividend-initiating ®rms experience a 3.9% increase in stock prices while dividend-omitting ®rms suer a 9.5% drop in stock prices. Similarly, Michaely et al. (1995) ®nd that the market's reaction to initiation is associated with a signi®cant price increase of 3.4%, while omissions register a price decrease of about 7%. Viewed collectively, these studies suggest that the dividend announcement eect is a re¯ection of the information signaled by managers through changes in the company's dividend policy. Despite the theoretical importance of dividend-signaling models, more recent empirical evidence calls into question the value of dividends as a signaling mechanism. Contrary to the conventional prediction of dividend-signaling models, both DeAngelo et al. (1996) and Benartzi et al. (1997) ®nd little support for the notion that dividends serve as a signaling mechanism to help investors distinguish among ®rms with superior earnings prospects. The ®nding of an earnings rebound following dividend reductions reported by
4
See, for example, Petit (1972), Charest (1978), Aharony and Swary (1980), Benesh et al. (1984), Dielman and Oppenheimer (1984), Ghosh and Woolridge (1988), Healy and Palepu (1988), Christie (1994), Dhillon and Johnson (1994), Sant and Cowan (1994) and Michaely et al. (1995).
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2075
DeAngelo et al. (1992) and Jensen and Johnson (1995), and the temporary decline in earnings following dividend omissions reported by Healy and Palepu (1988) seem to suggest that the signaling explanation does not fully capture all the information conveyed in the ®rm's dividend decision. Christie (1994) ®nding of a U-shaped pattern between the observed stock price response and the magnitude of dividend reduction/omission adds to the uncertainty concerning the information conveyed when a dividend is reduced or omitted. Based on this discussion, it may be inferred that the dividend decision alone may not be adequate to resolve the informational asymmetry between inside managers and outside investors. This gives rise to an important ®nancial implication: that the signaling explanation for dividend changes may not be complete, and signaling via dividends should be interpreted in conjunction with other activities in order to better understand the message conveyed by managers. In other words, dividend omission alone may not convey all information about the ®rm's dividend decision. Another mechanism may be at work in in¯uencing the market's evaluation of the ®rm's dividend action. Given the two bodies of literature, those relating to banking and to dividends, it may be inferred that the commonality between the two lies in the role of mitigating the costs of asymmetric information. It is possible that the ®rm's banking relationship may constitute an important signal of ®rm quality. Since banks are shown to play an important role in processing information and monitoring major corporate activities, and given that the dividend policy of ®rms represents a major corporate decision, it is conjectured that the presence of bank lending serves as a mechanism that in¯uences the market's evaluation of the dividend action. We believe the presence of bank debt may have a mitigating eect on the market's reaction to dividend omission announcements, something not provided by other forms of private or public debt. In eect, banks may serve as a unique form of monitor in the debt-for-dividend substitution argument of Jensen (1986). Further, given the evidence that the bene®ts of bank monitoring accrue primarily to small ®rms, we believe the mitigating eect of bank debt on the market's reaction to dividend omission is more pronounced for small ®rms than large ®rms. 3. Research design 3.1. Sampling procedure In identifying the potential omissions, this study follows the selection criteria employed by Christie (1994). The sample is collected by examining
2076
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
the dividend distribution record in the Center for Research in Security Prices (CRSP) Daily Master File. The actual announcement date of each omission is obtained by examining the Wall Street Journal Index for the years 1978±1996. The focus of the search is on ®rms that pay quarterly cash dividends in US currency, but excludes regulated ®rms. For ®rms that have more than one dividend omission, subsequent omissions are included only after resumption and continuous payment for at least 10 years (Healy and Palepu, 1988). Following Christie, a dividend omission is identi®ed when the elapsed time between the ex-dates of adjacent quarterly payments exceeds the ®rm's quarterly payment schedule by more than two months. For example, for ®rms to be included in the sample, the ex-dates of adjacent quarterly payments must be separated by at least six months, where the three months represent the implied quarterly schedule and the additional months represent the time lapse that could go beyond the three month implied schedule. The time interval of more than two months is imposed because the payment of quarterly dividends for some ®rms is unevenly spaced throughout the year. Similarly, a potential omission is also identi®ed if the time interval between the ®rm's last quarterly dividend and the month the ®rm ceases to be listed on CRSP or the time interval between the ®rm's last quarterly dividend and the end of the search period, December 1996, exceeds the ®rm's implied quarterly schedule by more than two months. The sample is then subjected to the following screening criteria: 1. The announcement date of dividend omission (t 0) must be reported in the Wall Street Journal. 2. There is no other ®rm-speci®c information that is announced on the day of, or the day preceding, the announcement of dividend omission. 3. There must be at least ®ve consecutive quarterly cash payments immediately before the announcement of dividend omission. (Christie (1994) imposed similar criteria to allow sucient time for dividend expectations to be formed.) 4. The dividends omitted must have been payable in US currency. 5. Firms involved in the process of merger or liquidation are excluded. 6. Firms that signal an intent to reduce or omit prior to the actual announcement are excluded. Firms must have return data available in the CRSP Daily Master File, and the ®rm's ®nancial information must be available in Standard & Poor's COMPUSTAT PC PLUS and annual reports. Data on managerial ownership of common stock must also be available in the corporate proxy statements or Value Line Investment Survey. Table 1 presents the screening criteria used to determine the ®nal sample based on all announcements identi®ed over the
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2077
Table 1 Criteria used to determine the sample of dividend omissions based on an initial sample of 901 announcements of dividend omission identi®ed over the period from 1 January 1978 through 31 December 1996 Dividend omissions Number of potential announcements Screening criteria Either announcement date or ®rm is not reported in the Wall Street Journal Announcement of other ®rm-speci®c events on days 0 or )1 Fewer than ®ve consecutive quarterly dividends immediately preceding the announcement Previous dividend payments are not in US currency Signal intention to reduce or omit prior to the announcement Undergoing merger or liquidation Information not available in COMPUSTAT, proxy statements or Value Line Investment Survey Insucient information in the annual reports Observations eliminated for statistical reasons or missing returns Total number of announcements eliminated Total number of usable announcements
901 297 216 26 6 14 52 32 7 21 671 230
study period. 5 Given that this study attempts to dierentiate the eect of bank monitoring on small and large ®rms that announce dividend omissions, we follow the market value of equity classi®cation scheme employed by Slovin et al. (1992) to separate large from small ®rms. 3.2. Statistical procedures A multi-part statistical methodology is employed. Standard event study methodology is ®rst used to calculate abnormal returns around announcements 5 There is one other factor that must be taken into account in developing a sample of dividend omissions. Kalay (1980) argues that involuntary reductions, such as those induced by debt restrictions, cannot be viewed as carrying the same signaling value as reductions that are voluntary. (The idea of voluntary management actions sending stronger signals than involuntary actions has undergone its own evolution in the literature; see, e.g., Cornett and Tehranian (1994).) Following Kalay, and to account for this possibility, we examined the debt covenant section of the Moody's Manuals series. Only six of our sample ®rms were found to suer forced omission due to covenant restriction. However, we found that investor response to the dividend action of those ®rms was statistically no dierent than the rest of the sample. In fact, those observations were spread evenly throughout the sample. We draw no conclusion from this observation other than to support Kalay's ®nding that forced dividend reductions are uncommon. Further, we are unable to distinguish whether banks are more strictly enforcing loan covenants or whether the covenants of banks are more restrictive by nature.
2078
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
of dividend omissions. Two-day abnormal returns (days )1 and 0) are computed for each ®rm in the sample. The market model parameters are estimated over the period day )316 to )67. 6 Mean two-day abnormal returns are calculated for the entire sample and the two subsamples. 7 To explain the abnormal returns observed at the time of dividend omission, several cross-sectional regression models are estimated for both small and large ®rms. The dependent variable in each model is the two-day abnormal return. Independent variables representing bank monitoring, together with variables that measure the monitoring eort of other types of lenders as well as control variables are employed to explain the announcement period return. These models are shown in the respective tables of results, and are explained below. 3.3. Description of independent variables 3.3.1. Proxy for bank debt The variable that proxies for the extent of the bank's monitoring eort is the level of the ®rm's outstanding bank debt, de®ned as the ®rm's bank debt in current liabilities, bank debt issued via bank lines of credit and revolving credit agreements, and other bank debt in long-term liabilities. While each of these entries is measured and included separately, a single variable representing the total of the three is employed in tests where no dierential impact in the variables would be hypothesized. Because there is no uniform reporting of detailed information about a ®rm's bank debt, this study follows Hirschey et al. (1990) and Slovin et al. (1990) in using a proxy for bank debt obtained from COMPUSTAT. This amount is then scaled by the market value of equity of the ®rm, which is de®ned as the closing share price as of the prior ®scal year-end, multiplied by the ®rm's common shares outstanding. The market value at ®scal year-end is obtained from COMPUSTAT. As in Slovin et al. (1990), this study uses bank debt issued via bank lines of credit and revolving credit agreements as an additional proxy for outstanding 6 The choice of the estimation period is based on ®ndings of Ghosh and Woolridge (1988) and Christie (1994). Both studies report that the cumulative abnormal return calculated over some period of time preceding the announcement of a dividend reduction or omission contributes signi®cantly in explaining the announcement period returns. Since a full trading quarter prior to the announcement (day )66 to )2) will be included as an independent variable in the multivariate regression models, the parameter estimation period ends on day )67. A 250 day estimation period represents one trading year. 7 The market model parameters were re-estimated using a procedure suggested by Dimson (1979) in an attempt to eliminate the bias in the estimation of beta caused by any potential problem of non-synchronous trading of small ®rm securities. The market model parameters were obtained by regressing the security returns against the lagged, matching and leading market returns. The results of this procedure altered neither the ®ndings nor conclusions of the study.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2079
debt. This information is obtained from the footnotes section of the ®rm's annual report for the ®scal year-end prior to the announcement of dividend omission. The data on outstanding bank lines of credit and revolving credit agreements are scaled by the ®rm's market value of equity. The variable ``other bank debt'' represents bank debt that does not fall into the above-mentioned categories. These data also are obtained from the notes to the ®rm's annual report for the ®scal year-end preceding the announcement of dividend omission. The variable representing other bank debt is also scaled by the ®rm's market value of equity. 3.3.2. Proxy variables for non-bank lending relationships Private non-bank debt, PRD, is de®ned as all other private long-term debt. These data are collected from the footnote section of the ®rm's annual report for the ®scal year-end prior to the announcement of dividend omission. The data are scaled by the ®rm's market value of equity. A variable representing public debt (PBD) is explicitly identi®ed in the ®rm's annual report under the notes on long-term debt/obligations, and also is scaled by the ®rm's market value of equity. 3.3.3. Control variables Three other independent variables, dividend yield (DY), the percentage of insider ownership of common stock (IN) and previous cumulative abnormal return (PCAR) are included in the model to serve as control variables. Dividend yield is de®ned as the most recent quarterly dividend prior to announcement of the dividend omission, divided by the share value two days prior to the announcement. This information is obtained from the CRSP Daily Master File. The fraction of equity held by inside managers is included in the crosssectional regression to control for the eect of information signaling inherent in managerial ownership. Managerial ownership is de®ned as the fraction of the ®rm's common stock owned by all executive ocers and directors as a group, and this information is collected from the ®rm's most recent proxy statement, or if not available, the most recent Value Line Investment Survey issued prior to the announcement of dividend omission. The cumulative abnormal return measured over the interval of day )66 to )2 is included in the cross-sectional regression to capture the eect of news releases or leakages occurring during the quarter leading up to the announcement of dividend omission. The inclusion of the PCAR variable has the additional bene®t of accounting for any potential shift in beta. Further, Firth (1996) shows that the announcement of dividend changes by one ®rm results in stock price reactions not only for the announcing ®rm but also for other nonannouncing ®rms within the same industry. The PCAR variable should capture
2080
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
any such industry eect, eliminating the need for a separate industry control variable.
4. Empirical results 4.1. Descriptive statistics Panel A of Table 2 contains descriptive statistics for the 100 small and 130 large ®rms. The mean market capitalization of small ®rms is $55.4 million, while the mean for large ®rms is $268.8 million. Outstanding bank debt averages $28.6 million for small ®rms and $77.5 for large ®rms. The other differences between small and large ®rms appear to be similarly scaled, and are signi®cantly dierent in all categories except ``other bank debt outstanding.'' Notice that the market reaction to dividend omission is signi®cantly less negative for small ®rms than for large ®rms. The dierence in the mean two-day CARs is 2.17% (t 1:715), which is signi®cant at the 0.05 level in a one-tailed test. This suggests that the eects of bank monitoring, as evidenced by the presence of a bank lending relationship, is of greater bene®t to small ®rms than large ®rms. Panel B of the table shows the variables scaled by the market value of equity. This shows that small ®rms and large ®rms with banking relationships are proportioned in similar fashion. 4.2. Cross-sectional regression results Tables 3 and 4 report the results of tests between small and large ®rms. A Chow test con®rms that the results for small ®rms are signi®cantly dierent from those of large ®rms, as an examination of the coecients would seem to indicate. Further, diagnostics (including White's test) show no evidence of heteroscedasticity or other econometric problems, therefore the tables are derived using ordinary least squares regression. 8 Table 3 shows the aggregated bank debt variables and other forms of debt outstanding. 9 For small ®rms, the combined bank debt variable (BD) is
8 The models were also run using total assets in the denominator instead of market value of equity. This was done for all debt categories to control for the fraction of each in the capital structure. The results did not vary from those using market value of equity. 9 The model was also run using the total amount of bank lines granted, as a measure of undrawn loan commitments. The results of those tests did not vary appreciably from those obtained using bank debt outstanding.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2081
Table 2 Descriptive statistics of small and large ®rms announcing dividend omissions over the sample period from 1 January 1978 through 31 December 1996a Variable
$000s Small ®rm (mean)
Panel A: Variables unscaled BDCL 5937 LINE 18,160 OBD 4466 BD 28,564 GRT 44,828 PRD 24,697 PBD 18,493 MKTVAL 55,443 TD/TA (%) CAR (%) PCAR (%)
30.92 )4.66 )6.49
Large ®rm (mean) 22,974 41,852 12,629 77,454 145,437 81,062 54,534 268,757 27.40 )6.83 )8.14
Mean dierence (t-statistic)
Prob > jT j
)2.940 )2.661 )1.592 )3.029 )3.808 )2.913 )2.705 )4.494
0.0038 0.0086 0.1130 0.0029 0.0002 0.0042 0.0075 0.0001
1.213 1.715 0.503
Panel B: Variables scaled by the market value of equity BDCL 14.42 15.64 )0.291 LINE 35.12 24.80 1.932 OBD 5.72 7.26 )0.690 BD 55.26 47.71 0.925 GRT 86.03 74.19 1.178 PRD 50.10 33.78 1.965 PBD 32.35 18.76 1.803
0.2270 0.0443 0.6153 0.7714 0.0549 0.4906 0.3560 0.2399 0.0514 0.0737
a The full sample of dividend omission consists of 230 announcements. The sample of small and large ®rms contain 100 and 130 announcements, respectively. All variables are measured as of the ®scal year-end prior to the omission announcements. BDCL is the bank debt in current liabilities, LINE is outstanding loans drawn under bank lines of credit and revolving credit agreements, OBD is other outstanding bank debt, BD is the total bank debt and equals BDCL + LINE + OBD, GRT equals the total amount of bank debt granted, PRD is private non-bank debt and PBD is public debt, MKTVAL is the market value of equity of the ®rm and is de®ned as the closing price for the ®scal year-end multiplied by the ®rm's common shares outstanding. The market value at ®scal yearend and BDCL were obtained from COMPUSTAT and all other variables were obtained from the ®rm's annual reports. TD/TA is total debt to total assets. CAR is the two-day abnormal return and PCAR is the cumulative abnormal return over the interval immediately preceding the dividend omission announcement, as explained in Section 3.3.3. * Denotes statistical signi®cance at the 0.01 level in two-tailed tests. ** Statistically signi®cant at the 0.05 level in a one-tail test. *** Denotes statistical signi®cance at the 0.10 level in two-tailed tests.
signi®cant at the 0.01 level (t 3:073). For large ®rms, both the coecients of the combined bank debt variable (BD) and private non-bank debt (PRD) are not statistically signi®cant. This suggests that the monitoring services provided by private lenders, be they banks or non-banks, do not constitute an important signal in aecting the way investors react to announcements of
2082
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
Table 3 Cross-sectional regression results for small and large ®rms announcing dividend omissions over the period from 1 January 1978 through 31 December 1996a;b;c Variable Constant BD PRD PBD DY IN PCAR
Small ®rms
Pr > jT j
Coecient
t-statistic
)5.2800 0.0351 0.0162 0.0008 )3.6636 0.0672 )0.0909
)3.990 3.073 1.505 0.086 )6.034 2.155 )3.372
F value 10.914 Prob > F 0.0001 Adjusted R2 0.3753 n 100
0.0001 0.0028 0.1357 0.9319 0.0001 0.0337 0.0011
Large ®rms
Pr > jT j
Coecient
t-statistic
)2.3165 )0.0149 )0.0117 )0.0423 )2.0589 0.0101 )0.0401
)2.148 )1.567 )0.782 )2.353 )3.489 0.336 )1.720
0.0337 0.1197 0.4357 0.0202 0.0007 0.7375 0.0879
F value 4.718 Prob > F 0.0002 Adjusted R2 0.1474 n 130
a
Testable proposition: Bank debt provides a stronger signal than private non-bank debt in mitigating the negative abnormal return associated with the announcement of dividend omission for small ®rms, but no such relationship exists for large ®rms. CAR a0 b1 BD b2 PRD b3 PBD b4 DY b5 IN b6 PCAR; where CAR is the two-day cumulative abnormal return associated with dividend omission, BD is the total bank debt (scaled by the market value of equity) and equals BDCL + LINE + OBD, BDCL is the bank debt in current liabilities (scaled by the market value of equity), LINE is outstanding loans drawn under bank lines of credit and revolving credit agreements (scaled by the market value of equity), OBD is other outstanding bank debt (scaled by the market value of equity), PRD is private non-bank debt (scaled by the market value of equity), PBD is public debt (scaled by the market value of equity), DY is dividend yield and equals the quarterly dividend immediately prior to the announcement divided by the share price on day )2, IN is the percentage of managerial ownership of common stock and PCAR is the cumulative abnormal return over the interval immediately preceding the dividend omission announcement, as explained in Section 3.3.3. b Diagnostic checks, including White's test, showed no evidence of heteroscedasticity or other econometric problems. c A Chow test shows that small ®rms dier from large ®rms, F value 4.436 (Prob > F 0.0211).
dividend omission by large ®rms. This ®nding is consistent with Diamond's (1991) view that larger, better-known ®rms need not rely on the monitoring services of banks, and may choose to borrow directly from the public debt market. In Table 4, the aggregated bank debt variable is broken down into components; bank debt in current liabilities, lines of credit and revolving credit agreements, and other bank debt in long-term liabilities. For small ®rms the short-term bank debt proxy, bank debt in current liabilities (BDCL), has signi®cant positive explanatory power (t 2:665). However, the other proxies for bank debt are not signi®cant. This ®nding implies that the extent of bank monitoring as measured by bank debt in current liabilities (BDCL) is viewed as
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2083
Table 4 Cross-sectional regression results for small and large ®rms announcing dividend omissions over the period from 1 January 1978 through 31 December 1996a;b;c Variable Constant BDCL LINE OBD PRD PBD DY IN PCAR
Small ®rms
Pr > jT j
Coecient
t-statistic
)5.4481 0.0757 0.0170 0.0500 0.0201 )0.0001 )3.7179 0.0698 )0.0867
)4.127 2.665 1.108 1.397 1.835 )0.009 )6.136 2.250 )3.216
F value 8.661 Prob > F 0.0001 Adjusted R2 0.3824 n 100
0.0001 0.0091 0.2708 0.1657 0.0698 0.9931 0.0001 0.0269 0.0018
Large ®rms
Pr > jT j
Coecient
t-statistic
)2.5873 )0.0334 0.0121 )0.0293 )0.0133 )0.0417 )1.9195 0.0065 )0.0357
)2.346 )2.251 0.644 )0.752 )0.889 )2.292 )3.089 0.214 )1.520
0.0206 0.0262 0.5211 0.4534 0.3760 0.0236 0.0025 0.8306 0.1311
F value 3.943 Prob > F 0.0004 Adjusted R2 0.1543 n 130
a
Testable proposition: Upon the announcement of dividend omission, the abnormal return experienced by small ®rms is positively related to the level of outstanding bank debt in the capital structure of the ®rm. For large ®rms, the relationship is non-positive. CAR a0 b1 BDCL b2 LINE b3 OBD b4 PRD b5 PBD b6 DY b7 IN b8 PCAR; where CAR is the two-day cumulative abnormal return associated with dividend omission, BDCL is the bank debt in current liabilities (scaled by the market value of equity), LINE is outstanding loans drawn under bank lines of credit and revolving credit agreements (scaled by the market value of equity), OBD is other outstanding bank debt (scaled by the market value of equity), PRD is private non-bank debt (scaled by the market value of equity), PBD is public debt (scaled by the market value of equity), DY is the dividend yield and equals the quarterly dividend immediately prior to the announcement divided by the share price on day )2, IN is the percentage of managerial ownership of common stock and PCAR is the cumulative abnormal return over the interval immediately preceding the dividend omission announcement, as explained in Section 3.3.3. b Diagnostic checks, including White's test, showed no evidence of heteroscedasticity or other econometric problems. c A Chow test shows that small ®rms dier from large ®rms, F value 3.711 (Prob > F 0.0178).
a suciently powerful signal, which is consistent with the notion that the bank's immediate concern is repayment of what is currently due. 10 The signi®cantly positive coecient of the bank monitoring variable BDCL suggests that the higher the level of bank debt, the less negative the stock price
10
When total current liabilities (CL) is substituted for BDCL, the CL variable is insigni®cant, indicating that BDCL is the relevant consideration, not the ®rm's overall near-term ®nances. When both variables are included in the regression, BDCL retains its signi®cance while CL shows none. We report the results using only BDCL because it is capable of standing alone and avoids any problems of multicollinearity.
2084
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
response to announcement of dividend omission. In other words, investors in small ®rms value the third-party monitoring role assumed by banks, indicating that investors are using this second signal when evaluating the ®rm's dividend decision. These results are not only consistent with other empirical evidence showing that the presence of a banking relationship constitutes an important signaling mechanism, but are also consistent with studies that indicate the bene®ts of bank monitoring accrue primarily to small ®rms. On the other hand, the results for large ®rms indicate the short-term bank debt proxy, bank debt in current liabilities (BDCL), is signi®cantly negative (t 2:251). This indicates that the more short-term bank debt the ®rm has, the more negative is the abnormal return associated with announcement of dividend omission. This is just the opposite of small ®rms. In this case, the presence of banks as monitors may be overshadowed by the gravity of the omission event, i.e., the ®nancial stress on the ®rm. With respect to other debt and control variables, the coecient associated with public debt is signi®cantly negative at the 0.05 level for large ®rms only. Like high levels of bank debt, high levels of public debt may be viewed as increasing the likelihood of ®nancial distress. Hence, the value of bank monitoring of larger, better-known ®rms appears minimal. The variable PCAR is signi®cantly negative for small ®rms and not signi®cant for large ®rms. This means that for small ®rms the more negative the abnormal return in the prior quarter, the less negative is the announcement period return. In other words, for small ®rms there is either some anticipation of the omission event in the preceding quarter, or the presence of bank monitoring in¯uences investor interpretation of new information and they adjust their expectations accordingly. This is not the case for large ®rms. While the price movement at the time of the omission announcement may be signi®cant for large ®rms, the statistical insigni®cance between periods indicates that there is no relationship between the preceding price action and the omission event. Given that other private non-bank lenders also obtain information and perform a monitoring function, we examine whether such lenders are aorded the same credibility as banks. As can be seen in both Tables 3 and 4, the coecient associated with private non-bank debt (PRD) is not statistically signi®cant for small ®rms. This ®nding implies that investors do not perceive the monitoring eorts of private non-bank lenders to be as important a signaling mechanism as bank lenders in assessing the value of the ®rm upon announcement of dividend omission. This ®nding is not surprising given the fact that private non-bank debt generally has a longer-term maturity than bank debt. For example, the mean and median maturities of bank debt in James's (1987) sample are 5.6 and 6 years, respectively, while the mean and median maturities of privately placed debt are 15.34 and 15 years, respectively. In this regard, as argued by Barclay and Smith (1995), most bank debt is short-term,
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2085
and the short-term nature of the debt contributes to making the monitoring activities of banks more eective. 5. Summary and conclusions This study investigates the role of bank monitoring in in¯uencing the market's response to a ®rm's decision to omit cash dividend payments. Results show that in the presence of a bank lending relationship, announcements of dividend omission by small ®rms are interpreted dierently than similar announcements made by large ®rms. More speci®cally, in the case of small ®rms the signal associated with the ®rm's banking relationship provides a framework for interpretation of the ®rm's dividend decision. Further, the presence of banks as monitors is observed to be of greater informational value to investors than the presence of private non-bank lenders. The results for large ®rms indicate that investors do not value the role of bank monitoring, and tend to react more negatively to announcements of dividend omission in the presence of high levels of bank debt. This response appears to be driven by concern over the ®rm's ability to meet its ®nancial obligations, rather than any third-party review of these better-known ®rms. In a broader context, by considering the impact of bank monitoring on the market's response to dividend announcements, this study makes two contributions. First, it extends the banking literature by expanding the bank's monitoring role to include inferences drawn from dividend theory. Further, by partitioning private debt into bank and non-bank sources, this study provides a better understanding of the dierential impact of lender monitoring on equity value. As its second contribution, the study adds to our understanding of whether banks, as monitors, add credibility to the dividend decisions of ®rms, and small ®rms in particular. Consistent with the theory of bank monitoring, it is shown that banks serve an important signaling role in in¯uencing investors' assessment of the dividend decisions of small ®rms.
References Aharony, J., Swary, I., 1980. Quarterly dividend and earnings announcements and stockholders' returns: An empirical analysis. Journal of Finance 35, 1±12. Barclay, M., Smith, C., 1995. The maturity structure of corporate debt. Journal of Finance 50, 609±631. Benartzi, S., Michaely, R., Thaler, R., 1997. Do changes in dividends signal the future or the past? Journal of Finance 52, 1007±1034. Benesh, G., Keown, A., Pinkerton, J., 1984. An examination of market reaction to substantial shifts in dividend policy. Journal of Financial Research 1984, 131±142.
2086
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
Bhattacharya, S., 1979. Imperfect information, dividend policy, and the `bird in the hand' fallacy. Bell Journal of Economics and Management Science 10, 259±270. Bhattacharya, S., Thakor, A., 1993. Contemporary banking theory. Journal of Financial Intermediation 3, 2±50. Billett, M., Flannery, M., Gar®nkel, J., 1995. The eect of lender identity on a borrowing ®rm's equity return. Journal of Finance 50, 699±718. Boyd, J., Prescott, E., 1986. Financial intermediary-coalitions. Journal of Economic Theory 38, 211±232. Charest, G., 1978. Dividend information, stock returns and market eciency: II. Journal of Financial Economics 6, 297±330. Christie, W., 1994. Are dividend omissions truly the cruelest cut of all? Journal of Financial and Quantitative Analysis 29, 459±480. Cornett, M., Tehranian, H., 1994. An examination of voluntary versus involuntary security issuances by commercial banks. Journal of Financial Economics 35, 99±122. DeAngelo, H., DeAngelo, L., Skinner, D., 1992. Dividends and losses. Journal of Finance 47, 1837±1863. DeAngelo, H., DeAngelo, L., Skinner, D., 1996. Reversal of fortune: Dividend signaling and the disappearance of sustained earnings growth. Journal of Financial Economics 40, 341±371. Dhillon, U., Johnson, H., 1994. The eect of dividend changes on stock and bond prices. Journal of Finance 49, 281±289. Diamond, D., 1984. Financial intermediation and delegated monitoring. Review of Economic Studies 51, 393±414. Diamond, D., 1991. Monitoring and reputation: The choice between bank loans and directly placed debt. Journal of Political Economy 99, 689±721. Dielman, T., Oppenheimer, H., 1984. An examination of investor behavior during periods of large dividend changes. Journal of Financial and Quantitative Analysis 19, 197±216. Dimson, E., 1979. Risk measurement when shares are subject to infrequent trading. Journal of Financial Economics 7, 197±226. Fama, E., 1985. What's dierent about banks? Journal of Monetary Economics 15, 29±39. Firth, M., 1996. Dividend changes, abnormal returns, and intra-industry ®rm valuations. Journal of Financial and Quantitative Analysis 31, 189±211. Ghosh, C., Woolridge, J.R., 1988. An analysis of shareholder reaction to dividend cuts and omissions. Journal of Financial Research 11, 281±294. Healy, P., Palepu, K., 1988. Earnings information conveyed by dividend initiations and omissions. Journal of Financial Economics 21, 149±175. Hirschey, M., Slovin, M., Zaima, J., 1990. Bank debt, insider trading and the return to corporate sellos. Journal of Banking and Finance 14, 85±98. James, C., 1987. Some evidence of the uniqueness of bank loans. Journal of Financial Economics 19, 217±235. James, C., Wier, P., 1990. Borrowing relationships, intermediation, and the cost of issuing public securities. Journal of Financial Economics 28, 149±171. Jensen, G., Johnson, J., 1995. The dynamics of corporate dividend reductions. Financial Management 24, 31±51. Jensen, M., 1986. Agency cost of free cash ¯ow, corporate ®nance, and takeovers. American Economic Review 76, 323±329. John, K., Williams, J., 1985. Dividends, dilution, and taxes: A signaling equilibrium. Journal of Finance 40, 1053±1070. Kalay, A., 1980. Signaling, information content, and the reluctance to cut dividends. Journal of Financial and Quantitative Analysis 15, 855±869. Leland, H., Pyle, D., 1977. Information asymmetries, ®nancial structure and ®nancial intermediaries. Journal of Finance 32, 371±387.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2087
Michaely, R., Thaler, R., Womack, K., 1995. Price reactions to dividend initiations and omissions: Overreaction or drift? Journal of Finance 50, 573±608. Miller, M., Rock, K., 1985. Dividend policy under asymmetric information. Journal of Finance 40, 1031±1052. Petit, R., 1972. Dividend announcements, security performance, and capital market eciency. Journal of Finance 27, 993±1007. Rajan, R., 1992. Insiders and outsiders: The choice between informed and arm's-length debt. Journal of Finance 47, 1367±1400. Ramakrishnan, R., Thakor, A., 1984. Information reliability and a theory of ®nancial intermediation. Review of Economic Studies 52, 415±432. Sant, R., Cowan, A., 1994. Do dividends signal earnings? The case of omitted dividends. Journal of Banking and Finance 18, 1113±1133. Slovin, M., Johnson, S., Glascock, J., 1992. Firm size and the information content of bank loan announcements. Journal of Banking and Finance 16, 1057±1071. Slovin, M., Sushka, M., Hudson, C., 1988. Corporate commercial paper, note issuance facilities, and shareholder wealth. Journal of International Money and Finance 7, 289±302. Slovin, M., Sushka, M., Hudson, C., 1990. External monitoring and its eect on seasoned common stock issues. Journal of Accounting and Economics 12, 397±417. Slovin, M., Young, J., 1990. Bank lending and initial public oerings. Journal of Banking and Finance 14, 729±740. Wansley, J., Elayan, F., Collins, M.C., 1993. Investment opportunities and ®rm quality: An empirical investigation of the information in bank lines of credit. Working paper, University of Tennessee. Williamson, S., 1986. Costly monitoring, ®nancial intermediation, and equilibrium credit rationing. Journal of Monetary Economics 18, 159±179.
The link between bank monitoring and corporate dividend policy: The case of dividend omissions Soo-Wah Low a, Louis Glorfeld b, Douglas Hearth b, James N. Rimbey b,* b
a National University of Malaysia, Kuala Lumpur, Malaysia Department of Finance, BA-302, Sam M. Walton College of Business Administration, University of Arkansas, Fayetteville, AR 72701, USA
Received 23 November 1998; accepted 26 October 2000
Abstract This study investigates whether bank monitoring in¯uences investor response to a borrowing ®rm's decision to omit its dividend payments. We establish a new link between the theories of banking and dividend policy in an examination of how bank monitoring and ®rm dividend signals complement one another to resolve information asymmetries. Results indicate that, for small ®rms, investors interpret the dividend decision as a function of bank monitoring and the dividend signals taken together. Also reported are the results of tests examining the dierences between the monitoring eects of banks versus public and private non-bank lenders. Ó 2001 Elsevier Science B.V. All rights reserved. JEL classi®cation: G14; G21; G35 Keywords: Bank monitoring; Dividend signaling; Dividend omission; Firm size
*
Corresponding author. Tel.: +501-575-4505; fax: +501-575-8407. E-mail address: [email protected] (J.N. Rimbey).
0378-4266/01/$ - see front matter Ó 2001 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 4 2 6 6 ( 0 0 ) 0 0 1 7 0 - 9
2070
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
1. Introduction The purpose of this study is to investigate whether bank monitoring in¯uences the market's response to a ®rm's decision to omit its cash dividend. In particular, this study establishes a new link between the theories of banking and dividend policy by examining how bank monitoring and dividends as signals may complement each other in mitigating information asymmetries. Information-based theories of ®nancial intermediation suggest that banks provide a unique information production and monitoring service in an informationally asymmetric capital market. 1 These theories imply that an imperfect capital market recognizes the value of bank monitoring and uses bank debt as a signal of ®rm value. A number of empirical studies show that the presence of a bank lending relationship disseminates information about the ®rm that in¯uences the market's assessment of major corporate decisions. 2 For example, studies have shown that the presence of bank debt aects the market assessment of corporate sello decisions, the issuance of commercial paper, and seasoned equity oerings. Other studies have demonstrated a relationship between bank debt, ®rm value, and the underpricing of initial public oerings. In short, the theoretical basis and empirical evidence suggest that bank debt serves as an important signal of ®rm quality and thus contributes to reducing information asymmetries. Research also suggests that the bene®ts of banks as information specialists and monitors of corporate activity may have the greatest eect on small ®rms (e.g., Fama, 1985; Slovin et al., 1992). Whereas much information may be available about large ®rms, providing them easier access to capital markets, small ®rms are less well known. Hence, their ability to obtain credit from banks may serve as a signal of value. It is in this context that we study the interaction between ®rm size and banking relationships. In previous studies of payout policy, various theories have been set forth to explain the relationship between dividends and ®rm value. Signaling theory suggests that in a market characterized by asymmetric information, dividends can be used as a signaling device to communicate private information. In general, dividend-signaling models posit that dividend announcements convey information about the ®rm's current and/or future cash ¯ows, therefore changes in the value of the ®rm around the time of dividend announcements should be proportionate to the unexpected change in dividend policy (e.g., Bhattacharya, 1979; Miller and Rock, 1985; John and Williams, 1985).
1
Much of this literature is reviewed in Bhattacharya and Thakor (1993). See, for example, Hirschey et al. (1990), James (1987), James and Wier (1990), Slovin et al. (1988), Slovin et al. (1990) and Slovin and Young (1990). 2
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2071
While a large body of empirical evidence documents a positive association between the announcement of dividend changes and ®rm value, more recent empirical evidence calls into question the value of dividends as signals. 3 This leads to the conjecture that the market reaction to a ®rm's dividend decision may be in¯uenced by other factors, such as the presence of banks as monitors of ®rm activity. To investigate the relationship between the market reaction to dividend omission announcements and bank debt, a sample of ®rms that omitted their dividends between 1978 and 1996 is identi®ed. Daily stock price returns for each ®rm in the sample are collected along with various measures of bank debt. The sample is further divided into subsamples consisting of large and small ®rms using the same classi®cation scheme as Slovin et al. (1992). A sequence of multivariate statistical tests is then conducted to examine the role of banks as lender/monitors. Our ®ndings show a strong positive relationship between bank lending and the market's assessment of dividend omission announcements for small ®rms. Further, the market's reaction to dividend omissions by small ®rms having high levels of bank debt is far less negative than it is when small ®rms have little or no bank debt. On the other hand, for large ®rms there appears to be little relationship between bank debt and the market response to dividend omission. Our ®ndings also suggest that bank debt has a greater mitigating impact on the market response to dividend omission than either non-bank private debt or public debt. The balance of this paper is organized as follows. Section 2 discusses the relationship between bank monitoring and dividend omission announcements. Section 3 describes the research design employed by this study. Section 4 presents our empirical ®ndings and Section 5 oers our conclusions. 2. Bank monitoring and dividend omissions 2.1. Information-based theory of ®nancial intermediation Leland and Pyle (1977) develop a signaling model that emphasizes asymmetric information as a primary reason to explain the existence of ®nancial intermediaries. They argue that ®nancial intermediaries can signal the
3 Examples of studies ®nding a positive association between dividend change announcements and ®rm value include Christie (1994), Healy and Palepu (1988) and Aharony and Swary (1980). Some of the recent studies questioning the value of dividends as a signaling mechanism include Benartzi et al. (1997), DeAngelo et al. (1996), Jensen and Johnson (1995) and Christie (1994).
2072
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
proprietary information of many borrowers at a cost that is lower than that of each individual borrower attempting to signal alone. In his model of delegated monitoring, Diamond (1984) extends Leland and Pyle by showing that private information about borrowers can be monitored more eciently by a ®nancial intermediary on behalf of its depositors (as suppliers of funds) rather than by many individual depositors acting as individual lenders. Diamond provides further insight by showing that ®nancial intermediaries possess a cost advantage in producing information, and that the monitoring function provided by intermediaries is cost eective in resolving the ®nancial contracting problems caused by asymmetric information. Rajan (1992) explores the costs and bene®ts of borrowing from informed versus arm's-length sources, and oers an explanation of why ®rms may seek alternatives to bank monitoring (and control) of ®rm investment decisions. Other models of ®nancial intermediaries as delegated monitors have been developed by Ramakrishnan and Thakor (1984), Boyd and Prescott (1986), and Williamson (1986). In summary, these theoretical models of ®nancial intermediation suggest that in a market with imperfect information, ®nancial intermediaries have a comparative advantage in collecting private information and monitoring the activities of borrowers relative to the direct monitoring by many individual lenders. Fama (1985) expands this notion by distinguishing banks from other types of ®nancial intermediaries and argues that banks are indeed a special type of ®nancial institution. He draws a distinction between inside and outside debt, and classi®es bank debt as inside debt, de®ned as a contract in which the creditor (the bank) gets access to the private information about the borrowing ®rm that is not available to other market participants. This con®dential disclosure of information by managers allows banks to know more about the quality of the ®rm than any other investor. As noted by Wansley et al. (1993), and Bhattacharya and Thakor (1993), bank debt provides a stronger signal of the ®rm's value than other forms of debt because bank lenders are better informed relative to other creditors. The bank's willingness to risk its ®nancial capital by granting loans to the ®rm re¯ects positive private information about the ®rm. There is a substantial amount of empirical evidence supporting the unique nature of bank loans. James (1987), for example, ®nds that borrowers experience a signi®cant positive abnormal return upon announcing new bank loan arrangements, versus a non-positive insigni®cant abnormal return for publicly placed debt and a signi®cant negative stock price response for debt placed privately with insurance companies. Billett et al. (1995) also ®nd that bank loan announcements elicit signi®cant positive abnormal returns to the announcing ®rms. They report signi®cant positive abnormal returns for both new and renewed loans, with renewed loans generating slightly larger stock price reactions, on average.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2073
2.2. The role of banks in monitoring managerial decisions Banks may also serve another monitoring function. Studies of various managerial actions have concluded that the existence and level of bank debt in¯uences market perceptions. In general these studies show that the market uses the banking relationship as a mechanism to certify the ®rm's actions. For example, Hirschey et al. (1990) ®nd signi®cant positive returns around corporate sello announcements by companies with high bank debt. By contrast, they ®nd little market reaction to sello announcements by companies with little or no bank debt. Slovin et al. (1990) ®nd that ®rms with extensive bank ®nancing experience virtually no stock price response to announcements of seasoned equity oerings. Firms with little bank debt experience signi®cant and negative stock price responses to seasoned equity oering announcements. The presence of banks as monitors described in these studies also ®ts well with the tenets of agency theory. In each of these cases the actions of managers generate free cash ¯ow, which could be invested sub-optimally (Jensen, 1986). In eect, market participants may be reassured by the bank's presence as a third-party monitor. 2.3. Bank monitoring and ®rm size Prior studies have suggested that the monitoring and evaluation functions of banks are more important for small companies than for larger ones. Fama (1985) argues that small ®rms incur lower contracting costs by securing ®nancing from banks as opposed to obtaining ®nancing from the public debt market, which entails a high cost of disclosure. He suggests that small ®rms ®nd it cheaper to borrow from banks since they are required to supply information to fewer lenders (the banks). Large ®rms, which have less of a ®nancial contracting problem, ®nd it more economical to distribute information more widely and obtain ®nancing from the public debt market. In his model of choice between bank loans and directly placed debt, Diamond (1991) argues that small companies with credit ratings that lie toward the middle of the spectrum rely heavily on bank loans since the quality of these ®rms cannot be easily veri®ed by outsiders. As such, these ®rms have an incentive to borrow from banks in order to obtain monitoring services and to acquire reputation through time by building up their credit histories. Once that reputation is established, these ®rms can then move into the public marketplace. Empirical evidence strongly suggests that the bene®ts of banks as information specialists and monitors of corporate activity accrue primarily to small ®rms. Slovin et al. (1992), for example, ®nd that the market's reaction to the renewal or initiation of bank loan arrangements is far more pronounced in small ®rms than it is in large ®rms.
2074
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2.4. Dividend omissions and signaling The notion of the ``informational content'' of dividends was modeled formally by Bhattacharya (1979), John and Williams (1985) and Miller and Rock (1985). The common theme of these models is that in the presence of asymmetric information between managers and outside investors, dividends can be used as a signaling mechanism by managers to communicate their assessment of the ®rm's current performance and future prospects. In general, dividendsignaling models posit that dividend announcements transmit information about the ®rm's future and/or current cash ¯ows and consequently the changes in the value of the ®rm around dividend announcements should be proportionate to the changes in dividend policy. Thus, when a ®rm unexpectedly increases (decreases) dividends, it signals management's optimistic (pessimistic) outlook for the future, and based on this information, outside investors should revise their expectations about the ®rm's value. There is a substantial amount of empirical evidence that documents a positive association between the announcement of dividend changes and stock price reaction. 4 There is general agreement among these studies that the market responds positively to favorable dividend changes (initiations, increases, and resumptions) and negatively to unfavorable dividend changes (reductions and omissions). However, the available empirical evidence also indicates that unfavorable dividend changes elicit market reactions that are much greater in magnitude than favorable dividend changes. For example, Healy and Palepu (1988) show that dividend-initiating ®rms experience a 3.9% increase in stock prices while dividend-omitting ®rms suer a 9.5% drop in stock prices. Similarly, Michaely et al. (1995) ®nd that the market's reaction to initiation is associated with a signi®cant price increase of 3.4%, while omissions register a price decrease of about 7%. Viewed collectively, these studies suggest that the dividend announcement eect is a re¯ection of the information signaled by managers through changes in the company's dividend policy. Despite the theoretical importance of dividend-signaling models, more recent empirical evidence calls into question the value of dividends as a signaling mechanism. Contrary to the conventional prediction of dividend-signaling models, both DeAngelo et al. (1996) and Benartzi et al. (1997) ®nd little support for the notion that dividends serve as a signaling mechanism to help investors distinguish among ®rms with superior earnings prospects. The ®nding of an earnings rebound following dividend reductions reported by
4
See, for example, Petit (1972), Charest (1978), Aharony and Swary (1980), Benesh et al. (1984), Dielman and Oppenheimer (1984), Ghosh and Woolridge (1988), Healy and Palepu (1988), Christie (1994), Dhillon and Johnson (1994), Sant and Cowan (1994) and Michaely et al. (1995).
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2075
DeAngelo et al. (1992) and Jensen and Johnson (1995), and the temporary decline in earnings following dividend omissions reported by Healy and Palepu (1988) seem to suggest that the signaling explanation does not fully capture all the information conveyed in the ®rm's dividend decision. Christie (1994) ®nding of a U-shaped pattern between the observed stock price response and the magnitude of dividend reduction/omission adds to the uncertainty concerning the information conveyed when a dividend is reduced or omitted. Based on this discussion, it may be inferred that the dividend decision alone may not be adequate to resolve the informational asymmetry between inside managers and outside investors. This gives rise to an important ®nancial implication: that the signaling explanation for dividend changes may not be complete, and signaling via dividends should be interpreted in conjunction with other activities in order to better understand the message conveyed by managers. In other words, dividend omission alone may not convey all information about the ®rm's dividend decision. Another mechanism may be at work in in¯uencing the market's evaluation of the ®rm's dividend action. Given the two bodies of literature, those relating to banking and to dividends, it may be inferred that the commonality between the two lies in the role of mitigating the costs of asymmetric information. It is possible that the ®rm's banking relationship may constitute an important signal of ®rm quality. Since banks are shown to play an important role in processing information and monitoring major corporate activities, and given that the dividend policy of ®rms represents a major corporate decision, it is conjectured that the presence of bank lending serves as a mechanism that in¯uences the market's evaluation of the dividend action. We believe the presence of bank debt may have a mitigating eect on the market's reaction to dividend omission announcements, something not provided by other forms of private or public debt. In eect, banks may serve as a unique form of monitor in the debt-for-dividend substitution argument of Jensen (1986). Further, given the evidence that the bene®ts of bank monitoring accrue primarily to small ®rms, we believe the mitigating eect of bank debt on the market's reaction to dividend omission is more pronounced for small ®rms than large ®rms. 3. Research design 3.1. Sampling procedure In identifying the potential omissions, this study follows the selection criteria employed by Christie (1994). The sample is collected by examining
2076
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
the dividend distribution record in the Center for Research in Security Prices (CRSP) Daily Master File. The actual announcement date of each omission is obtained by examining the Wall Street Journal Index for the years 1978±1996. The focus of the search is on ®rms that pay quarterly cash dividends in US currency, but excludes regulated ®rms. For ®rms that have more than one dividend omission, subsequent omissions are included only after resumption and continuous payment for at least 10 years (Healy and Palepu, 1988). Following Christie, a dividend omission is identi®ed when the elapsed time between the ex-dates of adjacent quarterly payments exceeds the ®rm's quarterly payment schedule by more than two months. For example, for ®rms to be included in the sample, the ex-dates of adjacent quarterly payments must be separated by at least six months, where the three months represent the implied quarterly schedule and the additional months represent the time lapse that could go beyond the three month implied schedule. The time interval of more than two months is imposed because the payment of quarterly dividends for some ®rms is unevenly spaced throughout the year. Similarly, a potential omission is also identi®ed if the time interval between the ®rm's last quarterly dividend and the month the ®rm ceases to be listed on CRSP or the time interval between the ®rm's last quarterly dividend and the end of the search period, December 1996, exceeds the ®rm's implied quarterly schedule by more than two months. The sample is then subjected to the following screening criteria: 1. The announcement date of dividend omission (t 0) must be reported in the Wall Street Journal. 2. There is no other ®rm-speci®c information that is announced on the day of, or the day preceding, the announcement of dividend omission. 3. There must be at least ®ve consecutive quarterly cash payments immediately before the announcement of dividend omission. (Christie (1994) imposed similar criteria to allow sucient time for dividend expectations to be formed.) 4. The dividends omitted must have been payable in US currency. 5. Firms involved in the process of merger or liquidation are excluded. 6. Firms that signal an intent to reduce or omit prior to the actual announcement are excluded. Firms must have return data available in the CRSP Daily Master File, and the ®rm's ®nancial information must be available in Standard & Poor's COMPUSTAT PC PLUS and annual reports. Data on managerial ownership of common stock must also be available in the corporate proxy statements or Value Line Investment Survey. Table 1 presents the screening criteria used to determine the ®nal sample based on all announcements identi®ed over the
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2077
Table 1 Criteria used to determine the sample of dividend omissions based on an initial sample of 901 announcements of dividend omission identi®ed over the period from 1 January 1978 through 31 December 1996 Dividend omissions Number of potential announcements Screening criteria Either announcement date or ®rm is not reported in the Wall Street Journal Announcement of other ®rm-speci®c events on days 0 or )1 Fewer than ®ve consecutive quarterly dividends immediately preceding the announcement Previous dividend payments are not in US currency Signal intention to reduce or omit prior to the announcement Undergoing merger or liquidation Information not available in COMPUSTAT, proxy statements or Value Line Investment Survey Insucient information in the annual reports Observations eliminated for statistical reasons or missing returns Total number of announcements eliminated Total number of usable announcements
901 297 216 26 6 14 52 32 7 21 671 230
study period. 5 Given that this study attempts to dierentiate the eect of bank monitoring on small and large ®rms that announce dividend omissions, we follow the market value of equity classi®cation scheme employed by Slovin et al. (1992) to separate large from small ®rms. 3.2. Statistical procedures A multi-part statistical methodology is employed. Standard event study methodology is ®rst used to calculate abnormal returns around announcements 5 There is one other factor that must be taken into account in developing a sample of dividend omissions. Kalay (1980) argues that involuntary reductions, such as those induced by debt restrictions, cannot be viewed as carrying the same signaling value as reductions that are voluntary. (The idea of voluntary management actions sending stronger signals than involuntary actions has undergone its own evolution in the literature; see, e.g., Cornett and Tehranian (1994).) Following Kalay, and to account for this possibility, we examined the debt covenant section of the Moody's Manuals series. Only six of our sample ®rms were found to suer forced omission due to covenant restriction. However, we found that investor response to the dividend action of those ®rms was statistically no dierent than the rest of the sample. In fact, those observations were spread evenly throughout the sample. We draw no conclusion from this observation other than to support Kalay's ®nding that forced dividend reductions are uncommon. Further, we are unable to distinguish whether banks are more strictly enforcing loan covenants or whether the covenants of banks are more restrictive by nature.
2078
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
of dividend omissions. Two-day abnormal returns (days )1 and 0) are computed for each ®rm in the sample. The market model parameters are estimated over the period day )316 to )67. 6 Mean two-day abnormal returns are calculated for the entire sample and the two subsamples. 7 To explain the abnormal returns observed at the time of dividend omission, several cross-sectional regression models are estimated for both small and large ®rms. The dependent variable in each model is the two-day abnormal return. Independent variables representing bank monitoring, together with variables that measure the monitoring eort of other types of lenders as well as control variables are employed to explain the announcement period return. These models are shown in the respective tables of results, and are explained below. 3.3. Description of independent variables 3.3.1. Proxy for bank debt The variable that proxies for the extent of the bank's monitoring eort is the level of the ®rm's outstanding bank debt, de®ned as the ®rm's bank debt in current liabilities, bank debt issued via bank lines of credit and revolving credit agreements, and other bank debt in long-term liabilities. While each of these entries is measured and included separately, a single variable representing the total of the three is employed in tests where no dierential impact in the variables would be hypothesized. Because there is no uniform reporting of detailed information about a ®rm's bank debt, this study follows Hirschey et al. (1990) and Slovin et al. (1990) in using a proxy for bank debt obtained from COMPUSTAT. This amount is then scaled by the market value of equity of the ®rm, which is de®ned as the closing share price as of the prior ®scal year-end, multiplied by the ®rm's common shares outstanding. The market value at ®scal year-end is obtained from COMPUSTAT. As in Slovin et al. (1990), this study uses bank debt issued via bank lines of credit and revolving credit agreements as an additional proxy for outstanding 6 The choice of the estimation period is based on ®ndings of Ghosh and Woolridge (1988) and Christie (1994). Both studies report that the cumulative abnormal return calculated over some period of time preceding the announcement of a dividend reduction or omission contributes signi®cantly in explaining the announcement period returns. Since a full trading quarter prior to the announcement (day )66 to )2) will be included as an independent variable in the multivariate regression models, the parameter estimation period ends on day )67. A 250 day estimation period represents one trading year. 7 The market model parameters were re-estimated using a procedure suggested by Dimson (1979) in an attempt to eliminate the bias in the estimation of beta caused by any potential problem of non-synchronous trading of small ®rm securities. The market model parameters were obtained by regressing the security returns against the lagged, matching and leading market returns. The results of this procedure altered neither the ®ndings nor conclusions of the study.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2079
debt. This information is obtained from the footnotes section of the ®rm's annual report for the ®scal year-end prior to the announcement of dividend omission. The data on outstanding bank lines of credit and revolving credit agreements are scaled by the ®rm's market value of equity. The variable ``other bank debt'' represents bank debt that does not fall into the above-mentioned categories. These data also are obtained from the notes to the ®rm's annual report for the ®scal year-end preceding the announcement of dividend omission. The variable representing other bank debt is also scaled by the ®rm's market value of equity. 3.3.2. Proxy variables for non-bank lending relationships Private non-bank debt, PRD, is de®ned as all other private long-term debt. These data are collected from the footnote section of the ®rm's annual report for the ®scal year-end prior to the announcement of dividend omission. The data are scaled by the ®rm's market value of equity. A variable representing public debt (PBD) is explicitly identi®ed in the ®rm's annual report under the notes on long-term debt/obligations, and also is scaled by the ®rm's market value of equity. 3.3.3. Control variables Three other independent variables, dividend yield (DY), the percentage of insider ownership of common stock (IN) and previous cumulative abnormal return (PCAR) are included in the model to serve as control variables. Dividend yield is de®ned as the most recent quarterly dividend prior to announcement of the dividend omission, divided by the share value two days prior to the announcement. This information is obtained from the CRSP Daily Master File. The fraction of equity held by inside managers is included in the crosssectional regression to control for the eect of information signaling inherent in managerial ownership. Managerial ownership is de®ned as the fraction of the ®rm's common stock owned by all executive ocers and directors as a group, and this information is collected from the ®rm's most recent proxy statement, or if not available, the most recent Value Line Investment Survey issued prior to the announcement of dividend omission. The cumulative abnormal return measured over the interval of day )66 to )2 is included in the cross-sectional regression to capture the eect of news releases or leakages occurring during the quarter leading up to the announcement of dividend omission. The inclusion of the PCAR variable has the additional bene®t of accounting for any potential shift in beta. Further, Firth (1996) shows that the announcement of dividend changes by one ®rm results in stock price reactions not only for the announcing ®rm but also for other nonannouncing ®rms within the same industry. The PCAR variable should capture
2080
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
any such industry eect, eliminating the need for a separate industry control variable.
4. Empirical results 4.1. Descriptive statistics Panel A of Table 2 contains descriptive statistics for the 100 small and 130 large ®rms. The mean market capitalization of small ®rms is $55.4 million, while the mean for large ®rms is $268.8 million. Outstanding bank debt averages $28.6 million for small ®rms and $77.5 for large ®rms. The other differences between small and large ®rms appear to be similarly scaled, and are signi®cantly dierent in all categories except ``other bank debt outstanding.'' Notice that the market reaction to dividend omission is signi®cantly less negative for small ®rms than for large ®rms. The dierence in the mean two-day CARs is 2.17% (t 1:715), which is signi®cant at the 0.05 level in a one-tailed test. This suggests that the eects of bank monitoring, as evidenced by the presence of a bank lending relationship, is of greater bene®t to small ®rms than large ®rms. Panel B of the table shows the variables scaled by the market value of equity. This shows that small ®rms and large ®rms with banking relationships are proportioned in similar fashion. 4.2. Cross-sectional regression results Tables 3 and 4 report the results of tests between small and large ®rms. A Chow test con®rms that the results for small ®rms are signi®cantly dierent from those of large ®rms, as an examination of the coecients would seem to indicate. Further, diagnostics (including White's test) show no evidence of heteroscedasticity or other econometric problems, therefore the tables are derived using ordinary least squares regression. 8 Table 3 shows the aggregated bank debt variables and other forms of debt outstanding. 9 For small ®rms, the combined bank debt variable (BD) is
8 The models were also run using total assets in the denominator instead of market value of equity. This was done for all debt categories to control for the fraction of each in the capital structure. The results did not vary from those using market value of equity. 9 The model was also run using the total amount of bank lines granted, as a measure of undrawn loan commitments. The results of those tests did not vary appreciably from those obtained using bank debt outstanding.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2081
Table 2 Descriptive statistics of small and large ®rms announcing dividend omissions over the sample period from 1 January 1978 through 31 December 1996a Variable
$000s Small ®rm (mean)
Panel A: Variables unscaled BDCL 5937 LINE 18,160 OBD 4466 BD 28,564 GRT 44,828 PRD 24,697 PBD 18,493 MKTVAL 55,443 TD/TA (%) CAR (%) PCAR (%)
30.92 )4.66 )6.49
Large ®rm (mean) 22,974 41,852 12,629 77,454 145,437 81,062 54,534 268,757 27.40 )6.83 )8.14
Mean dierence (t-statistic)
Prob > jT j
)2.940 )2.661 )1.592 )3.029 )3.808 )2.913 )2.705 )4.494
0.0038 0.0086 0.1130 0.0029 0.0002 0.0042 0.0075 0.0001
1.213 1.715 0.503
Panel B: Variables scaled by the market value of equity BDCL 14.42 15.64 )0.291 LINE 35.12 24.80 1.932 OBD 5.72 7.26 )0.690 BD 55.26 47.71 0.925 GRT 86.03 74.19 1.178 PRD 50.10 33.78 1.965 PBD 32.35 18.76 1.803
0.2270 0.0443 0.6153 0.7714 0.0549 0.4906 0.3560 0.2399 0.0514 0.0737
a The full sample of dividend omission consists of 230 announcements. The sample of small and large ®rms contain 100 and 130 announcements, respectively. All variables are measured as of the ®scal year-end prior to the omission announcements. BDCL is the bank debt in current liabilities, LINE is outstanding loans drawn under bank lines of credit and revolving credit agreements, OBD is other outstanding bank debt, BD is the total bank debt and equals BDCL + LINE + OBD, GRT equals the total amount of bank debt granted, PRD is private non-bank debt and PBD is public debt, MKTVAL is the market value of equity of the ®rm and is de®ned as the closing price for the ®scal year-end multiplied by the ®rm's common shares outstanding. The market value at ®scal yearend and BDCL were obtained from COMPUSTAT and all other variables were obtained from the ®rm's annual reports. TD/TA is total debt to total assets. CAR is the two-day abnormal return and PCAR is the cumulative abnormal return over the interval immediately preceding the dividend omission announcement, as explained in Section 3.3.3. * Denotes statistical signi®cance at the 0.01 level in two-tailed tests. ** Statistically signi®cant at the 0.05 level in a one-tail test. *** Denotes statistical signi®cance at the 0.10 level in two-tailed tests.
signi®cant at the 0.01 level (t 3:073). For large ®rms, both the coecients of the combined bank debt variable (BD) and private non-bank debt (PRD) are not statistically signi®cant. This suggests that the monitoring services provided by private lenders, be they banks or non-banks, do not constitute an important signal in aecting the way investors react to announcements of
2082
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
Table 3 Cross-sectional regression results for small and large ®rms announcing dividend omissions over the period from 1 January 1978 through 31 December 1996a;b;c Variable Constant BD PRD PBD DY IN PCAR
Small ®rms
Pr > jT j
Coecient
t-statistic
)5.2800 0.0351 0.0162 0.0008 )3.6636 0.0672 )0.0909
)3.990 3.073 1.505 0.086 )6.034 2.155 )3.372
F value 10.914 Prob > F 0.0001 Adjusted R2 0.3753 n 100
0.0001 0.0028 0.1357 0.9319 0.0001 0.0337 0.0011
Large ®rms
Pr > jT j
Coecient
t-statistic
)2.3165 )0.0149 )0.0117 )0.0423 )2.0589 0.0101 )0.0401
)2.148 )1.567 )0.782 )2.353 )3.489 0.336 )1.720
0.0337 0.1197 0.4357 0.0202 0.0007 0.7375 0.0879
F value 4.718 Prob > F 0.0002 Adjusted R2 0.1474 n 130
a
Testable proposition: Bank debt provides a stronger signal than private non-bank debt in mitigating the negative abnormal return associated with the announcement of dividend omission for small ®rms, but no such relationship exists for large ®rms. CAR a0 b1 BD b2 PRD b3 PBD b4 DY b5 IN b6 PCAR; where CAR is the two-day cumulative abnormal return associated with dividend omission, BD is the total bank debt (scaled by the market value of equity) and equals BDCL + LINE + OBD, BDCL is the bank debt in current liabilities (scaled by the market value of equity), LINE is outstanding loans drawn under bank lines of credit and revolving credit agreements (scaled by the market value of equity), OBD is other outstanding bank debt (scaled by the market value of equity), PRD is private non-bank debt (scaled by the market value of equity), PBD is public debt (scaled by the market value of equity), DY is dividend yield and equals the quarterly dividend immediately prior to the announcement divided by the share price on day )2, IN is the percentage of managerial ownership of common stock and PCAR is the cumulative abnormal return over the interval immediately preceding the dividend omission announcement, as explained in Section 3.3.3. b Diagnostic checks, including White's test, showed no evidence of heteroscedasticity or other econometric problems. c A Chow test shows that small ®rms dier from large ®rms, F value 4.436 (Prob > F 0.0211).
dividend omission by large ®rms. This ®nding is consistent with Diamond's (1991) view that larger, better-known ®rms need not rely on the monitoring services of banks, and may choose to borrow directly from the public debt market. In Table 4, the aggregated bank debt variable is broken down into components; bank debt in current liabilities, lines of credit and revolving credit agreements, and other bank debt in long-term liabilities. For small ®rms the short-term bank debt proxy, bank debt in current liabilities (BDCL), has signi®cant positive explanatory power (t 2:665). However, the other proxies for bank debt are not signi®cant. This ®nding implies that the extent of bank monitoring as measured by bank debt in current liabilities (BDCL) is viewed as
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2083
Table 4 Cross-sectional regression results for small and large ®rms announcing dividend omissions over the period from 1 January 1978 through 31 December 1996a;b;c Variable Constant BDCL LINE OBD PRD PBD DY IN PCAR
Small ®rms
Pr > jT j
Coecient
t-statistic
)5.4481 0.0757 0.0170 0.0500 0.0201 )0.0001 )3.7179 0.0698 )0.0867
)4.127 2.665 1.108 1.397 1.835 )0.009 )6.136 2.250 )3.216
F value 8.661 Prob > F 0.0001 Adjusted R2 0.3824 n 100
0.0001 0.0091 0.2708 0.1657 0.0698 0.9931 0.0001 0.0269 0.0018
Large ®rms
Pr > jT j
Coecient
t-statistic
)2.5873 )0.0334 0.0121 )0.0293 )0.0133 )0.0417 )1.9195 0.0065 )0.0357
)2.346 )2.251 0.644 )0.752 )0.889 )2.292 )3.089 0.214 )1.520
0.0206 0.0262 0.5211 0.4534 0.3760 0.0236 0.0025 0.8306 0.1311
F value 3.943 Prob > F 0.0004 Adjusted R2 0.1543 n 130
a
Testable proposition: Upon the announcement of dividend omission, the abnormal return experienced by small ®rms is positively related to the level of outstanding bank debt in the capital structure of the ®rm. For large ®rms, the relationship is non-positive. CAR a0 b1 BDCL b2 LINE b3 OBD b4 PRD b5 PBD b6 DY b7 IN b8 PCAR; where CAR is the two-day cumulative abnormal return associated with dividend omission, BDCL is the bank debt in current liabilities (scaled by the market value of equity), LINE is outstanding loans drawn under bank lines of credit and revolving credit agreements (scaled by the market value of equity), OBD is other outstanding bank debt (scaled by the market value of equity), PRD is private non-bank debt (scaled by the market value of equity), PBD is public debt (scaled by the market value of equity), DY is the dividend yield and equals the quarterly dividend immediately prior to the announcement divided by the share price on day )2, IN is the percentage of managerial ownership of common stock and PCAR is the cumulative abnormal return over the interval immediately preceding the dividend omission announcement, as explained in Section 3.3.3. b Diagnostic checks, including White's test, showed no evidence of heteroscedasticity or other econometric problems. c A Chow test shows that small ®rms dier from large ®rms, F value 3.711 (Prob > F 0.0178).
a suciently powerful signal, which is consistent with the notion that the bank's immediate concern is repayment of what is currently due. 10 The signi®cantly positive coecient of the bank monitoring variable BDCL suggests that the higher the level of bank debt, the less negative the stock price
10
When total current liabilities (CL) is substituted for BDCL, the CL variable is insigni®cant, indicating that BDCL is the relevant consideration, not the ®rm's overall near-term ®nances. When both variables are included in the regression, BDCL retains its signi®cance while CL shows none. We report the results using only BDCL because it is capable of standing alone and avoids any problems of multicollinearity.
2084
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
response to announcement of dividend omission. In other words, investors in small ®rms value the third-party monitoring role assumed by banks, indicating that investors are using this second signal when evaluating the ®rm's dividend decision. These results are not only consistent with other empirical evidence showing that the presence of a banking relationship constitutes an important signaling mechanism, but are also consistent with studies that indicate the bene®ts of bank monitoring accrue primarily to small ®rms. On the other hand, the results for large ®rms indicate the short-term bank debt proxy, bank debt in current liabilities (BDCL), is signi®cantly negative (t 2:251). This indicates that the more short-term bank debt the ®rm has, the more negative is the abnormal return associated with announcement of dividend omission. This is just the opposite of small ®rms. In this case, the presence of banks as monitors may be overshadowed by the gravity of the omission event, i.e., the ®nancial stress on the ®rm. With respect to other debt and control variables, the coecient associated with public debt is signi®cantly negative at the 0.05 level for large ®rms only. Like high levels of bank debt, high levels of public debt may be viewed as increasing the likelihood of ®nancial distress. Hence, the value of bank monitoring of larger, better-known ®rms appears minimal. The variable PCAR is signi®cantly negative for small ®rms and not signi®cant for large ®rms. This means that for small ®rms the more negative the abnormal return in the prior quarter, the less negative is the announcement period return. In other words, for small ®rms there is either some anticipation of the omission event in the preceding quarter, or the presence of bank monitoring in¯uences investor interpretation of new information and they adjust their expectations accordingly. This is not the case for large ®rms. While the price movement at the time of the omission announcement may be signi®cant for large ®rms, the statistical insigni®cance between periods indicates that there is no relationship between the preceding price action and the omission event. Given that other private non-bank lenders also obtain information and perform a monitoring function, we examine whether such lenders are aorded the same credibility as banks. As can be seen in both Tables 3 and 4, the coecient associated with private non-bank debt (PRD) is not statistically signi®cant for small ®rms. This ®nding implies that investors do not perceive the monitoring eorts of private non-bank lenders to be as important a signaling mechanism as bank lenders in assessing the value of the ®rm upon announcement of dividend omission. This ®nding is not surprising given the fact that private non-bank debt generally has a longer-term maturity than bank debt. For example, the mean and median maturities of bank debt in James's (1987) sample are 5.6 and 6 years, respectively, while the mean and median maturities of privately placed debt are 15.34 and 15 years, respectively. In this regard, as argued by Barclay and Smith (1995), most bank debt is short-term,
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2085
and the short-term nature of the debt contributes to making the monitoring activities of banks more eective. 5. Summary and conclusions This study investigates the role of bank monitoring in in¯uencing the market's response to a ®rm's decision to omit cash dividend payments. Results show that in the presence of a bank lending relationship, announcements of dividend omission by small ®rms are interpreted dierently than similar announcements made by large ®rms. More speci®cally, in the case of small ®rms the signal associated with the ®rm's banking relationship provides a framework for interpretation of the ®rm's dividend decision. Further, the presence of banks as monitors is observed to be of greater informational value to investors than the presence of private non-bank lenders. The results for large ®rms indicate that investors do not value the role of bank monitoring, and tend to react more negatively to announcements of dividend omission in the presence of high levels of bank debt. This response appears to be driven by concern over the ®rm's ability to meet its ®nancial obligations, rather than any third-party review of these better-known ®rms. In a broader context, by considering the impact of bank monitoring on the market's response to dividend announcements, this study makes two contributions. First, it extends the banking literature by expanding the bank's monitoring role to include inferences drawn from dividend theory. Further, by partitioning private debt into bank and non-bank sources, this study provides a better understanding of the dierential impact of lender monitoring on equity value. As its second contribution, the study adds to our understanding of whether banks, as monitors, add credibility to the dividend decisions of ®rms, and small ®rms in particular. Consistent with the theory of bank monitoring, it is shown that banks serve an important signaling role in in¯uencing investors' assessment of the dividend decisions of small ®rms.
References Aharony, J., Swary, I., 1980. Quarterly dividend and earnings announcements and stockholders' returns: An empirical analysis. Journal of Finance 35, 1±12. Barclay, M., Smith, C., 1995. The maturity structure of corporate debt. Journal of Finance 50, 609±631. Benartzi, S., Michaely, R., Thaler, R., 1997. Do changes in dividends signal the future or the past? Journal of Finance 52, 1007±1034. Benesh, G., Keown, A., Pinkerton, J., 1984. An examination of market reaction to substantial shifts in dividend policy. Journal of Financial Research 1984, 131±142.
2086
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
Bhattacharya, S., 1979. Imperfect information, dividend policy, and the `bird in the hand' fallacy. Bell Journal of Economics and Management Science 10, 259±270. Bhattacharya, S., Thakor, A., 1993. Contemporary banking theory. Journal of Financial Intermediation 3, 2±50. Billett, M., Flannery, M., Gar®nkel, J., 1995. The eect of lender identity on a borrowing ®rm's equity return. Journal of Finance 50, 699±718. Boyd, J., Prescott, E., 1986. Financial intermediary-coalitions. Journal of Economic Theory 38, 211±232. Charest, G., 1978. Dividend information, stock returns and market eciency: II. Journal of Financial Economics 6, 297±330. Christie, W., 1994. Are dividend omissions truly the cruelest cut of all? Journal of Financial and Quantitative Analysis 29, 459±480. Cornett, M., Tehranian, H., 1994. An examination of voluntary versus involuntary security issuances by commercial banks. Journal of Financial Economics 35, 99±122. DeAngelo, H., DeAngelo, L., Skinner, D., 1992. Dividends and losses. Journal of Finance 47, 1837±1863. DeAngelo, H., DeAngelo, L., Skinner, D., 1996. Reversal of fortune: Dividend signaling and the disappearance of sustained earnings growth. Journal of Financial Economics 40, 341±371. Dhillon, U., Johnson, H., 1994. The eect of dividend changes on stock and bond prices. Journal of Finance 49, 281±289. Diamond, D., 1984. Financial intermediation and delegated monitoring. Review of Economic Studies 51, 393±414. Diamond, D., 1991. Monitoring and reputation: The choice between bank loans and directly placed debt. Journal of Political Economy 99, 689±721. Dielman, T., Oppenheimer, H., 1984. An examination of investor behavior during periods of large dividend changes. Journal of Financial and Quantitative Analysis 19, 197±216. Dimson, E., 1979. Risk measurement when shares are subject to infrequent trading. Journal of Financial Economics 7, 197±226. Fama, E., 1985. What's dierent about banks? Journal of Monetary Economics 15, 29±39. Firth, M., 1996. Dividend changes, abnormal returns, and intra-industry ®rm valuations. Journal of Financial and Quantitative Analysis 31, 189±211. Ghosh, C., Woolridge, J.R., 1988. An analysis of shareholder reaction to dividend cuts and omissions. Journal of Financial Research 11, 281±294. Healy, P., Palepu, K., 1988. Earnings information conveyed by dividend initiations and omissions. Journal of Financial Economics 21, 149±175. Hirschey, M., Slovin, M., Zaima, J., 1990. Bank debt, insider trading and the return to corporate sellos. Journal of Banking and Finance 14, 85±98. James, C., 1987. Some evidence of the uniqueness of bank loans. Journal of Financial Economics 19, 217±235. James, C., Wier, P., 1990. Borrowing relationships, intermediation, and the cost of issuing public securities. Journal of Financial Economics 28, 149±171. Jensen, G., Johnson, J., 1995. The dynamics of corporate dividend reductions. Financial Management 24, 31±51. Jensen, M., 1986. Agency cost of free cash ¯ow, corporate ®nance, and takeovers. American Economic Review 76, 323±329. John, K., Williams, J., 1985. Dividends, dilution, and taxes: A signaling equilibrium. Journal of Finance 40, 1053±1070. Kalay, A., 1980. Signaling, information content, and the reluctance to cut dividends. Journal of Financial and Quantitative Analysis 15, 855±869. Leland, H., Pyle, D., 1977. Information asymmetries, ®nancial structure and ®nancial intermediaries. Journal of Finance 32, 371±387.
S.-W. Low et al. / Journal of Banking & Finance 25 (2001) 2069±2087
2087
Michaely, R., Thaler, R., Womack, K., 1995. Price reactions to dividend initiations and omissions: Overreaction or drift? Journal of Finance 50, 573±608. Miller, M., Rock, K., 1985. Dividend policy under asymmetric information. Journal of Finance 40, 1031±1052. Petit, R., 1972. Dividend announcements, security performance, and capital market eciency. Journal of Finance 27, 993±1007. Rajan, R., 1992. Insiders and outsiders: The choice between informed and arm's-length debt. Journal of Finance 47, 1367±1400. Ramakrishnan, R., Thakor, A., 1984. Information reliability and a theory of ®nancial intermediation. Review of Economic Studies 52, 415±432. Sant, R., Cowan, A., 1994. Do dividends signal earnings? The case of omitted dividends. Journal of Banking and Finance 18, 1113±1133. Slovin, M., Johnson, S., Glascock, J., 1992. Firm size and the information content of bank loan announcements. Journal of Banking and Finance 16, 1057±1071. Slovin, M., Sushka, M., Hudson, C., 1988. Corporate commercial paper, note issuance facilities, and shareholder wealth. Journal of International Money and Finance 7, 289±302. Slovin, M., Sushka, M., Hudson, C., 1990. External monitoring and its eect on seasoned common stock issues. Journal of Accounting and Economics 12, 397±417. Slovin, M., Young, J., 1990. Bank lending and initial public oerings. Journal of Banking and Finance 14, 729±740. Wansley, J., Elayan, F., Collins, M.C., 1993. Investment opportunities and ®rm quality: An empirical investigation of the information in bank lines of credit. Working paper, University of Tennessee. Williamson, S., 1986. Costly monitoring, ®nancial intermediation, and equilibrium credit rationing. Journal of Monetary Economics 18, 159±179.