Printer Friendly

Bank Relationships and Firm Profitability.

Steven Ongena [*]

This paper examines how bank relationships affect firm performance. An empirical implication of recent theoretical models is that firms maintaining multiple bank relationships are less profitable than their single-bank peers. We investigate this empirical implication using a data set containing virtually all Norwegian publicly listed firms for the period 1979-1995. We find a robust and economically relevant negative two-way correspondence between the number of relationships and sales profitability. We also find that firms replacing a single relationship are on average smaller and younger than those firms choosing not to replace a single relationship.

Although practitioners and business analysts have long recognized the importance of bank relationships for firms (e.g., Stancill, 1980), academic interest in the topic has been rekindled by a slate of recent theoretical models. [1] This paper contributes to the growing literature on bank relationships by empirically examining the link between the firm's choice of the number of bank relationships and firm profitability.

Firms with valuable proprietary information may use fewer creditors in order to prevent information leakage (Bhattacharya and Chiesa, 1995). At the same time, financially distressed firms facing credit rationing at their main bank may be forced to seek additional and more costly bank finance elsewhere. Over time, weaker firms may switch banks more often, enticed by less informed outside offers promising lower interest rates (von Thadden, 1998). Hence, high quality firms can be conjectured 1) to maintain a single bank relationship and 2) to refrain from switching. The ongoing bilateral relationship may reinforce their sales profitability through easier access to credit and guaranteed confidentiality, though part of the firms' profits may flow to the inside bank through higher interest rates.

To test these conjectures, we study a data set that contains bank relationship and firm-specific data for most publicly listed Norwegian firms over a sixteen-year period. We control for firm characteristics such as age, size, debt structure, asset intangibility, and Tobin's Q; also, we estimate a two-equation model featuring sales profitability and the number of bank relationships as endogenous variables. There is a robust negative correspondence between these two variables. Ceteris paribus, firms with a bilateral bank relationship show a higher profitability. Firms that are more profitable maintain a bilateral relationship. Also. firms single relationship are, on average, smaller and younger than firms maintaining a single relationship throughout the sample years.

The remainder of the paper is organized as follows. Section I highlights the relevant literature linking the number of bank-firm relationships to firm profitability. Section II describes the data and the empirical specifications. Section III reports the results and Section IV concludes.

I. Bank Relationships and Firm Profitability

Recent theoretical models argue that close relationships between banks and firms may improve access to financing for firms, create value, and ultimately, improve firm performance. Bank relationships may widen contracting flexibility ex ante (Boot and Thakor, 1994), reduce agency problems through enhanced control (Rajan, 1992), and generally improve the availability of capital. For example, Petersen and Rajan (1995) model the interdependency between a firm's possibilities to borrow and the market power of the inside banks. They show that borrowing from banks with large market power facilitates inter-temporal sharing of rent surplus, while competition may hinder such accommodating policies.

A. Benefits of Close Single Bank Relationships

Close relationships may also enable reputation-building as a means for establishing enough credibility to eventually borrow through public debt or equity markets (Diamond, 1991), in effect, reducing the relative need for bank funds (Miarka, 1999). The confidentiality of bank lending may further limit the leakage of proprietary information (Campbell, 1979). Legal concerns, loss of reputation, or the negative impact on the performance of an important client may keep a bank from relaying confidential information obtained in an exclusive relationship, directly or through dispensed advice, to product market competitors of the firm. [2]

Confidentiality may be valuable for firms engaging in research and development (Bhattacharya and Chiesa, 1995), particularly for high-quality innovators. Yosha (1995), for example, argues that, if multilateral financing is more costly, then low-quality innovators may use such financing arrangements to credibly reveal the low value of their projects to the incumbent firm. On the other hand, von Rheinbaben and Ruckes (1998) note that multilateral banking entails not only higher transaction costs, but also more competitive interest rates.

Hence, they contend that the monotonic "negative" relationship between ex post profitability and the number of financing sources may be reversed. In addition, they argue that the innovator can control whether or not to disclose confidential information during the loan granting process.

B. Benefits of Multiple Relationships

A bank's ability to privately observe proprietary information and maintain a close relationship with its customer also creates a holdup problem. Fischer (1990), Sharpe (1990), and von Thadden (1998) argue that bank relationships arise even in a competitive loan market, because the incumbent bank has the ability to offer only loans above-cost to its best customers and "hold up" customers from receiving competitive financing elsewhere. The "inside bank" gains this monopoly power through its informational advantage over "outside" banks. A high-quality firm that attempts to switch to a competing uninformed bank gets pooled with low-quality firms and offered an even worse, break-even interest rate. Holdup costs are also present in Rajan (1992), since in his model, the bank has the power to withdraw financing when it perceives the firm to be inadequately managed. This degree of control can be costly because it reduces the incentives of the firm manager to exert effort.

One seemingly simple solution to the holdup problem is to maintain multiple bank relationships. For example, in von Thadden (1992), multiple bank relationships abate the informational lock-in problem, which reduces the interest rate charged by the inside banks. Multiple relationships may entail additional benefits for the firm. For example, while an exclusive relationship may positively affect current credit availability, it also exposes the firm to the risk that future credit needed to reinvest in the project may be withheld if the inside bank was affected by a negative liquidity shock (Detragiache, Garella, and Guiso, 2000). Firms may want to diversify this risk by maintaining multiple bank relationships.

Further, exclusive bank-firm ties could adversely shape management decisions and firm performance. For example, Bolton and Scharfstein (1996) argue that maintaining multiple creditors creates inefficient renegotiations, which may deter strategic default. But while beneficial ex ante, such a strategy may be costly ex post if firm distress is caused by exogenous liquidity shocks. Hence, all else equal, their model predicts that low default risk firms tend to borrow from more creditors. On the other hand (and this may especially be the case for large firms), banks may want to diversify firm-specific credit risk, thus resulting in more creditors for high default risk firms.

Multiple bank relationships may go hand-in-hand with lower credit availability, especially for distressed firms, as well as lower interest rates, less stringent bank controls, and (in case of information disclosure) lower sales profitability. Ultimately, then, determining the correspondence between the number of credit relationships and the firm's bottom line remains an empirical issue. But, as yet, we are unaware of any other paper that has tested this association directly.

C. Empirical Evidence on Relationships and Credit Terms

Most studies focus on the impact of the exclusivity of a relationship on credit availability and interest rates. For example, using data from different US surveys of small business financing, Petersen and Rajan (1994), Cole (1998), and more recently Scott (2000), find that a close relationship with a single institutional creditor increases the availability of credit. Analogously, Harhoff and Korting (1998b) and Angelini, Di Salvo, and Ferri (1998) document that credit availability for small German and Italian firms decreases with the number of relationships. On the other hand, Weinstein and Yafeh (1998) find that Japanese "main" bank clients enjoy superior access to capital resources.

Some evidence also broadly supports the potential negative impact of the number of relationships on the interest rate paid by the firm. For example, Weinstein and Yafeh (1998) document that (between 1977 and 1986) Japanese "main" banks, which typically nurture close and exclusive ties with their corporate clients, extracted higher interest payments than other banks. Angelini, et al. (1998) and Machauer and Weber (1999) find that a lower number of relationships result in higher interest rates for, respectively, small Italian and large German firms. However, these results have not been replicated for other samples and countries. For example, Harhoff and Korting (1998b) find that the number of relationships has no effect on interest rates in their analysis of a survey covering small German firms, while Petersen and Rajan (1994) find that, for small US firms, multiple relationships may even increase lending rates. Degryse and Van Cayseele (2000) report that Belgian firms taking loans from a secondary bank pay hi gher rates on these loans.

D. Empirical Evidence on Relationships and Firm Performance

Recent studies look at the impact of the banking arrangement on firm performance. For example, Weinstein and Yafeh (1998) ascertain that Japanese "main" banks imposed conservative investment policies on their clients, which inhibited growth and depressed firm profitability. Their results indicate a positive relationship between the number of creditors and firm profitability, if main bank clients in their sample have fewer other credit sources, ceteris paribus. Kang, Shivdasani, and Yamada (2000), on the other hand, find that close ties with healthy Japanese banks facilitated investment policies, enhancing shareholder wealth especially for firms with poor investment opportunities. Miarka (2000) also documents conflicting results with respect to the impact of different measures of relationship intensity on firm profitability (which is measured by ordinary income over sales).

Gorton and Schmid (2000) study the influence of German universal banks on the performance of German firms. Bank control rights from equity ownership improve firm performance, measured by the market-to-book value of equity, beyond what nonbank block holders can attain. Since "house "bank arrangements are often more exclusive (Elsas and Krahnen, 1998 and Machauer and Weber, 1999), this result seems to suggest a negative correspondence between the number of credit relationships and firm performance. [3]

Other papers focus on the firm's choice regarding the financing arrangement. For example, Detragiache, et al. (2000) find that larger and less profitable firms serviced by large banks are less likely to have a single banking relationship. They interpret the latter variable, measured in their paper by the "average size of the lending banks," as a proxy for bank liquidity. In their theoretical model, bank liquidity negatively affects the likelihood of a single banking equilibrium. Alternatively, this result could indicate that firms need additional relationships to counter being held up by a large main bank, or that firms with extended credit need to first "graduate" to a large main bank before engaging additional banks.

Other empirical work has emphasized the correspondence between firm distress and the number of bank relationships. For example, Harhoff and Korting (l998a) and Foglia, Laviola, and Marullo Reedtz (1998) find that financially distressed firms (in Germany and Italy respectively) have significantly more creditors than other, comparable non-distressed firms. This empirical evidence, though for smaller and medium sized firms, seems to indicate that the banks' desire for diversification may force firms (by rationing credit) to seek additional lenders.

In contrast to both sets of studies, we look more directly at the two-way correspondence between the number of bank relationships and firm profitability. We find that firms with a bilateral relationship are more profitable and that more profitable firms more often maintain a bilateral relationship. This result corresponds to the main intuition embedded in Yosha (1995) to the extent that observable profitability correlates with unobservable firm quality. If banks actively seek out highly profitable firms, in effect decreasing the cost of establishing an additional relationship for such firms, then one would expect profitable firms to have more relationships. In addition, our results extend the intuition about the banks' desire for diversification (possibly resulting in credit rationing) to a sample including larger and healthier firms. In most specifications, highly leveraged firms choose a single relationship. This possibly suggests that firms judged fit to carry a lot of debt do not face credit rationing by their single bank. The remainder of the paper develops these results.

II. Data and Empirical Specification

We analyze data first used by Ongena and Smith (2000b) in their study of the duration of bank relationships. Their data set contains observations of "primary" relationships that are gleaned from Kierulf's Handbook for most publicly listed, non-financial Norwegian companies between 1979 and 1995. The Norwegian setting (and this data set in particular) is well suited to explore the connection between firm profitability and the character of the financing arrangement. Around 90% of all commercial debt of Norwegian firms is financed by either a bank or a non-bank financial intermediary (see, for example, Statistical Yearbook of No Norway, 1996). Relationships are reported yearly, and only reported relationships typically involve both short- and long-term credit. In addition, the effect of switching a relationship on interest payments and firm profitability can be gleaned adequately from accounting statements, as bank debt is particularly important.

In any given year, between 65 and 75% of the publicly listed Norwegian firms report bilateral (i.e., single, "primary" bank relationships) in sharp contrast with most other European countries where bilateral relationships between medium- and large-sized firms and banks are often the exception (Ongena and Smith, 2000c). Furthermore, the banking sector also remained highly concentrated throughout the sample period. [4] Hence, the distinction between bilateral and multilateral financing may be sharp. Adding or ending a relationship with one of the few large banks will substantially affect the average firm in the sample. [5]

Each year, on average, 110 firms that are listed on the Oslo Stock Exchange register up to a maximum of four "primary" bank relationships. Listing and delisting activity throughout the sample period results in 235 different firms in the sample. The average firm reports at least one relationship during eight consecutive years that amounts to 1897 firm-year observations (Table I). The median firm in the sample uses only one "primary" bank, but around a quarter of the firms maintain multiple relationships (17% of the firms employ two banks, 7% employ three, and 2% of the firms report the maximum of four banks). This presence of firms with multiple relationships results in 2436 relationship-year observations, which are around 30% higher than the number of firm years.

When each relationship year is matched with firm-specific accounting information from the computer readable database FINLIS (accounting data is missing for a few firms and in the beginning of the sample period), and observations with extreme negative levels of sales profitability are removed, 1659 usable relationship-year observations are left. [6] Table II lists selected summary statistics for the remaining 1284 companies. Firms quite often alter their set of relationships. After matching, the sample contains 87 "events." In 30 cases, a firm replaces its single relationship. Firms also change (15 cases) or end (24 cases) one of their multiple relationships, or start an extra new relationship (18 cases).

Are firms using bilateral financing more profitable in sales than those firms using multilateral financing? To investigate this hypothesis directly, we start by simply regressing a measure of firm profitability on a dummy (dRELATION) equal to one (zero) when a firm maintains multiple (single) relationships. [7] The measure of firm profitability employed as dependent variable on the left-hand side is gross ROA (i.e., the percentage ratio of earnings before interest and taxes to the sum of the market value of equity, plus the book value of debt; see Table II).

The empirical specification further includes measures of firm size, debt structure, age, goodwill, other asset intangibility, and Tobin's Q as independent variables. First, we explain why these firm characteristics should enter the estimated reduced form. Firm size is included to control for the firm's market power and efficiency. Ceteris paribus, more market power, as well as greater efficiency may result in higher profitability. Size is measured by the log of the end-of-year sales and deflated by the Norwegian CPI (SALES).

The debt structure of a firm may affect firm profitability, although theory points in different directions with respect to its impact (Harris and Raviv, 1991). Firms either may use more debt to create tax shields (Modigliani and Miller, 1963), mitigate agency problems (Jensen, 1986; Stulz, 1990; and Ross, 1977), or use less debt to avoid external finance costs (Myers and Majluf, 1984). Debt also may influence strategic interaction among competitors, customers, and/or suppliers. For example, Brander and Lewis (1986) argue that firms commit to using a more aggressive product market strategy by choosing positive debt levels. Similarly, Titman (1984) shows that firms producing common products may be expected to have more debt as a commitment device--in such cases, shareholders will be reluctant to liquidate the firm. On the other hand, Hubert (1998) asserts that equity, and not debt, enhances the firm's competitiveness in the product market. In sum, the resultant effect of debt on profitability seems indetermina nt. The debt structure of the firm is measured as the ratio of the book value of debt to the sum of the market value of equity, plus the book value of debt (DEBT).

Firm age captures the length of the firm's track record. More profitable firms are more likely to survive, and older firms may have lower communication costs--hence, higher profitability in signaling models. Firm age is measured by the log of the age of the firm relative to its founding date (AGE). Asset intangibility, in general, proxies for firm opaqueness, while goodwill, in particular, may result from higher firm profitability. We use, as measures of asset intangibility, both the ratio of the total book value of goodwill and other intangibles to the sum of the market value of equity, plus the book value of debt (GOODWILL, OTHER INTANGIBLES). Finally, Tobin's Q proxies for the firm's investment opportunities. It is defined as the ratio of the end-of-year market value of equity plus book value of debt divided by the book value of assets (Q).

III. Results

We start by documenting the direct impact of bilateral versus multilateral financing on firm profitability in a parsimonious, static specification. Subsequently, we turn to a system of two equations in which profitability and the banking arrangement(s) are jointly determined. Also, we incorporate the dynamic effect of substituting bank relationship(s) on firm profitability, and motivate the introduction of time and industry effects. Next, we subject this system to a variety of robustness checks and then interpret the results. Finally, we investigate in more detail both the profitability of firms replacing a single relationship, and the switching decision itself.

A. Static Specification

We start by estimating the specification discussed so far, containing ROA as the dependent variable, and a constant, SALES, DEBT, AGE, GOODWILL, OTHER INTANGIBLES, Q, and dRELATION on the right-hand side. The first column in Table III (Model III.1) reports the results from OLS estimation. The coefficient on dRELATION is negative and significant at a 0.01 level; hence, firms with bilateral financing arrangements are more profitable. Maintaining multiple relationships decreases ROA by an economically significant 3.87%, which amounts to more than 40% of its standard deviation. Therefore, the form of banking arrangement seems important in explaining firm profitability.

The coefficient on DEBT is negative and significant at a 0.01 level, consistent with arguments focusing on external finance costs or more product market aggressiveness. The coefficient on SALES is significantly positive, because market power and efficiency may result in higher sales profitability. The coefficients ofAGE and GOODWILL are both positive and significant. The latter coefficient may reflect the higher propensity of more profitable firms to acquire other firms. The coefficient on Q is not significant.

B. Two Equations: Linear and Logit

Model III.1 assumes independence between the error term and the explanatory variables. First, the firm's choice of the number of relationships may depend upon its sales profitability. For example, a low-rated firm may have to draw upon multiple banks whose willingness to lend is constrained, but who may be willing to risk small stakes. Second, while public observation of the firm's number of relationships may reasonably be expected to precede the reporting and publication in Kieruif's Handbook, it is possible that this is not the case for all firms. Notice that reporting to Kierulf's occurs before firms release the annual reports. Publication of Kierulf's occurs after this release. In addition, profitability figures refer to the whole calendar year, while the observation of(a change in) the number of relationships (and the incumbent's reaction in a signaling model) will occur at some point during the "recording year" employed by Kierulf's.

To address interdependence, we estimate a model containing an equation with ROA as the dependent variable. A second equation has dRELATION as the dependent variable. The latter choice variable is binary (i.e., either a bilateral or a multilateral financing arrangement is chosen); hence, we employ a logit specification. [8] In the absence of cross-equation restrictions, estimating both equations jointly is equivalent to estimating the linear and the logit equation separately using two-stage least squares. Hence, we first run ROA and dRELATION on the set of exogenous variables using, respectively, a linear and logit model. Then, in a second stage, we use the fitted values of either dependent variable as a right-hand-side variable in the "other" equation.

We assume that contemporary changes in relationships do not determine the currently chosen banking arrangement. Also, we introduce two additional variables determining the choice of the number of relationships. BANKSHARE is defined as the (maximum) proportion of all registered bank connections accounted for by the firm's bank(s). The second variable, dDnBorCBK, takes the value of one if the firm has a relationship with either DnB (Bergens Bank or DnC before their merger) or CBK, and zero otherwise. We find justification for employing the first variable because firms having a relationship with a larger bank do not require additional banks to satisfy larger needs for credit and other financial services. In addition, firms may consider the number of possible competitors already serviced by the firm's bank(s) when determining their financing arrangement. The second variable is complementary. DnB and CBK are Norway's two largest banks, offering specialized credit and foreign exchange products that no other domest ic bank offers. In addition, these banks could be considered less subject to liquidity shocks (Detragiache, et al. 2000) or "too-big-to-fail."

C. Events and Effects

Model III.1 also assumes that firm profitability would be affected immediately by the firm's choice with respect to the number of relationships. But an immediate effect may not occur due to adjustment costs and time lags. Hence, analyzing the impact of changes in the number of relationships on firm profitability may provide an additional (although weaker) test of the impact of the number of creditors on firm profitability. We will look at the impact of the adjustments in the financial arrangement up to two years after the event. This time window seems a priori reasonable, and is commonly found in other types of firm performance studies (for example, Ber, Yafeh, and Yosha, 2000). Nevertheless, we will also report robustness tests with respect to this particular choice.

We define a dummy variable dSTART (dEND) to take the value of one in the first year or up to two years after a firm starts (ends) one of multiple relationships. In most cases, starting a new relationship involves moving from a bilateral to a multilateral banking arrangement, while ending a relationship results in a move in the opposite direction. A dummy variable dSINGLE takes the value of one in the first year or up to two years after a firm replaces a single relationship. The dummy variable dREPLACE takes the value of one in the first year or up to two years after a firm replaces one of multiple relationships. We do not consider a change in relationship due to a bank merger to be an event.

Finally, Model III.1 implicitly assumes that each firm chooses its size, debt structure, age, asset intangibility, Q, financial arrangement, corresponding profitability every year independent of its own choices in previous years, and independent of the contemporaneous choices made by all other firms. However, intertemporal links seem likely for most of these variables. Hence, it seems fruitful to introduce both year and firm effects. Computational constraints prevent the introduction of both year and firm effects in the logit model (see Greene, 1995, p. 435 for details); hence, we restrict ourselves to year and industry effects. Unfortunately, a "SIC/NAISC" type of classification scheme is not available in Norway. We resort to a classification scheme employed in Finansavisen, a business daily, from which we draw upon the 1995 edition. [9]

D. Results

Second stage results for both the linear and the logit equations are reported for Model 2 in Table III. The results are very interesting. While the coefficient on dRELATION halves in absolute value to -- 1.63, it remains significant at a 0.01 level, and continues to explain about 20% of the standard deviation of ROA. We list unadjusted standard errors, but the required adjustments are very small. [10] This is the case in all further reported specifications. Coefficients on firm characteristics are relatively unchanged from the single-equation fixed-effects model, except GOODWILL, which doubles in size. The coefficients on dREPLACE and dSTART are negative and significant, indicating that switching one of multiple relationships, or starting a new relationship is negatively related to the firm's profitability. On the other hand, the coefficient on dSINGLE is positive and marginally significant. The coefficient on dEND is not significant.

We also report estimated coefficients for the logit model. Less profitable, larger, more leveraged, and younger firms will establish more relationships. In addition, firms having a large main bank are more likely to have multiple relationships. The effects are also economically relevant. For example, an increase in ROA from 3 to 4% decreases the probability of having a multilateral relationship by around 1%. [11] An increase in leverage from 60 to 70% would increase the probability by around 2.5%. Doubling sales or halving age (at the mean) increases the likelihood of maintaining multiple relationships by around 5% each. Having a large main bank with a 40%, rather than a 30% market share, increases the likelihood by around 10%.

E. Fama-MacBeth

Next, we check the structural stability of this base specification for each consecutive year. The rationale for this additional step is that the Norwegian financial landscape altered dramatically during the sample period. Before deregulation in the mid-80s, chronic excess demand for credit cemented close and long-term relationships between borrowers and their banks (Drees and Pazarbasioglu, 1995). Deregulation led to credit expansion, feverish competition for market share, loan losses, a financial crisis, and eventually the effective nationalization of the entire banking sector in 1991.

In the spirit of Fama and MacBeth (1973) and Fama and French (1997), we estimate year-by-year two-equation models. Given the smaller number of observations (around 100 per year) and to prevent collinearity and convergence problems, we eliminate time and industry effects, event dummies, bank size variables, and the instrumenting of the return variable. We calculate averages and standard deviations, and report the results in Panel B.1 in the Appendix (we report standard deviations divided by the square root of the number of observations for easier comparison with the standard errors reported in the other models). The average coefficients of these 15 separate regressions and their significance levels (using standard Z-tests) are very similar to the results reported in the previous specification. Leverage is correlated negatively with firm profitability, while size, goodwill, other asset intangibility, and maintenance of a single relationship are positively correlated with firm profitability. In addition, large, young, low-growth, and less profitable firms seem to maintain more relationships.

The 15 coefficients on dRELATION are relatively stable--most are negative, and many are significant. [12] Nevertheless, there is an interesting decrease in the coefficient over time (from around -2.5% in the first four years of the sample period to about -7.5% in the last four years). The Norwegian banking crisis and the coincident increase in concentration may provide a potential explanation. For example, the effect on firm profitability of information leakage via multiple relationships may increase in concentration in the banking market, as each additional bank commands a significant market share. The correlation between the 15 coefficients and a Herfindahl-Hirschman Index for the banking market is indeed high: -0.51. Alternatively, depressed firm profitability during the crisis years may have resulted in credit rationing and multiple banking. Correlation between the coefficients and mean gross ROA or the Oslo Stock Exchange Index are 0.22 and -0.58, respectively. Nevertheless, given the small number of ob servations, we find these results merely suggestive.

F. Further Robustness Checks

Next, we subject the estimated results in Model III.2 to a battery of robustness checks. We start by varying the impact period of the adjustments in the financial arrangement. The Appendix in Panel A reports three additional impact "shadows": {0}, {0,1}, and {0,l,2,3}. For easy comparison, we repeat the results for the event window utilized to this point: {0,l,2}. We suppress results for most other coefficients because these are hardly affected by the induced changes in the definition of the event dummies. It is interesting to note that changes in the banking arrangement take two years before they have a significant effect on ROA. Replacing one of many relationships and starting a new relationship precedes a drop in profitability. Switching a bilateral relationship positively affects ROA but only marginally, and for one event window.

Also, we derive estimates for specifications using alternative measures of firm profitability. We employ the ratio of "pre-tax income" (earnings before interest and taxes, minus total interest expenses) to assets (PRETAX), and the ratio of earnings before interest and taxes to sales (PROFITS). The fixed costs of establishing relationship(s) may be part of the interest expenses. This implies that PRETAX may be an appropriate profitability measure. On the other hand, interest expenses may be higher for firms held up in a bilateral relationship. The second suggested variable (i.e., sales profitability) is attractive as a measure of firm profitability, since theoretical models directly linking the financing arrangement and firm profitability are grounded in product market behavior. This measure is somewhat less commonly used and, by its construction, prone to more variability and the occurrence of outliers.

The results for both alternative profitability measures are reported in Panel B in the Appendix. Overall results are broadly unaffected, with a few exceptions. In the Model B.2 (employing PRETAX as a return measure), the coefficient on dRELATION is no longer significant. However, it does remain negative and economically relevant. In the logit model, the coefficient on the profitability measures becomes substantially larger in absolute size. For example, an increase in PROFITS from 6 to 7% increases the likelihood of a bilateral relationship by about 9%. In addition, the coefficient on BANKSHARE and DEBT turn negative in the logit model. The latter coefficient will be negative in all further reported specifications. The positive sign in the base model is possibly due to the positive effect of financial distress on DEBT (as the market value of equity decreases) and dRELATION (as there may be credit rationing from the existing bank). Indeed, when we exclude firms with a Tobin's Q [less than] 1 from the sample ( 1327 observations remain), the coefficient on DEBT becomes -8.40 and significant at a 1% level. Alternatively, when we redefine DEBT in the base model by taking the ratio of the book value of debt to the sum of the book value of equity, plus the book value of debt, its coefficient becomes insignificant, while the other results are unaffected.

Also, we employ net income to book equity (Return On Equity), truncated at -100% and + 1000% (results are not reported). Again, overall results are broadly unaffected, except for the estimated coefficient on dRELATION, which equals -4.52 (which, with a standard error of 3.50, is no longer significantly different from zero at standard significance levels). But ROE is often noisy and driven by accounting and tax practices. In addition, interest expenses may be higher for single-bank firms, abating the negative correpondence between profitability and the number of relationships.

Next, we check our empirical model by using alternative definitions of firm size, as we are concerned that either non-linearities or the variable definition could partly determine the results reported so far. Hence, in Model III.2 we replace SALES alternately with each of the following: the log of the number of employees, the log of deflated firm assets, the log of deflated sales and the log of deflated sales squared, and dummies capturing below and above median deflated sales. Results are virtually unaffected. We further replace BANKSHARE and dDnBorCBK by BANKSIZE, which is defined as the ratio of the (maximum) asset size of the firm's bank(s) and assets. Lack of bank accounting data for earlier years in the sample limits the sample to 877 observations, and we are forced to drop industry effects, event dummies, and the instrumentation of ROA. Results are reported in the Appendix, Model B.4. The coefficient on BANKSIZE is insignificant, but other results are broadly unchanged.

To control better for dynamics we next add lagged sales profitability as an exogeneous variable to the base specification (Model III.2). We also add ownership concentration (the proportion of firm equity owned by the ten largest shareholders), because more concentrated ownership may improve sales profitability (for example, Ber, et al. 2000). Data on ownership concentration is missing for some firms in the sample, and only 1047 observations remain. Results are reported in Model B.5. Both added variables are positively and significantly related to the number of relationships, but the coefficient on OWNERSHIP CONCENTRATION becomes equal to -2.38 and significant at a 0.05 level in the ROA regression. Lagged sales determine current ROA positively. Otherwise, results are broadly unaffected.

Finally, analyzing relationship-year observations and the events affecting these relationships increases the representation in the sample of firms with multiple relationships and/or those subject to many events. In addition, events could be correlated through time. For example, once a firm replaces a bank, the probability may increase that another replacement will subsequently occur. To tackle both problems, a not-reported exercise considers only one relationship per firm, either the first relationship affected by an event, or the relationship ranked first by the firm (in case none were affected). While the sample drops to 1009 observations, the results are largely unaffected.

G. Yosha (1995) and von Rheinbaben and Ruckes (1998)

Smaller and "more tangible" firms, and possibly highly leveraged and older firms appear to be less profitable and engage fewer banks. Yet, ceteris paribus, firms with a bilateral relationship are more profitable; profitable firms more often maintain a bilateral relationship. This correspondence is robust and economically meaningful. For example, the difference between single and multiple bank firms in gross return on assets is between 3 to 6% in the reported models; an increase in gross ROA from 2 to 3% increases the probability of having a bilateral relationship by 1 to 10%.

In addition, we find that replacing one of many relationships is followed within two years by a significant and substantial decrease in firm profitability of around 4 to 7%. However, starting a new relationship decreases profitability by about 1.5%. Ending a relationship and switching a bilateral relationship seems to have a zero or moderately positive impact on profitability, although the latter results may be partly driven by the higher profitability observed for single bank firms.

All results discussed so far seem broadly in line with an ad-hoc dynamic extension of the Yosha (1995) model. Multilateral financing leaks information to the competitor and yields lower profitability. Starting a new relationship and replacing an existing relationship will increase the number of established competitors that can receive information about the innovator's quality. [13]

An interesting feature in Yosha (1995), although currently not reflected in our empirical model, is the interaction between opaqueness (AGE, GOODWILL, and OTHER INTANGIBLES) and the financing arrangement. In Yosha, firms with more proprietary information have a stronger relationship between profitability and the number of relationships. Introducing the interaction terms in the logit model is technically impossible; [14] hence, we replace dRELATION by its count equivalent, RELATION, defined as the number of relationships reported by the firm. Also, we replace the logit by a negative binomial model [15] and control for right censoring (firms can report only up to four relationships). In the second stage, we introduce in both equations the interaction terms with the fitted values of RELATION. Results are reported in Model 3 of Table III. [16] Only the coefficient on AGE * RELATION is significant at 5%. However, the latter coefficient is relatively small. When an additional relationship is added, a ten-year-old firm will decrease its ROA by 4.30%. On the other hand, a 20-year-old firm will see its ROA drop by a mere 0.14%.

Finally, we turn to a specification closer in spirit to the intuition embedded in von Rheinbaben and Ruckes (1998), who extend Yosha (1995). In their model, firms with an exante high credit rating will not disclose confidential information, but will maintain many relationships. We attempt to capture this possible non-monotonicity in the relationship between profitability and the number of bank connections by adding the fitted value of (RELATION) [2] to specification III.3 (not reported). The coefficient on this added variable is not significant.

H. Von Thadden (1998)

Next, we empirically investigate whether weaker firms are more likely to switch banks, thus leading to a Winner's Curse problem. Table IV takes a closer look at the characteristics of firms replacing a single relationship in the event year vis-a-vis the characteristics of all other single bank firms across their listing period. [17] We observe 30 switches of single relationships in the sample. As reported in the table, switching firms are, on average, smaller and younger in the switching year than other single bank firms across their listing period. They also seem less profitable, but the difference is not significant at standard significance levels. Hence, "bad" firms seem, on average, to switch more often than "good" ones, broadly confirming one implication of von Thadden (1998). Table IV also compares firms with bilateral and multiple relationships (i.e., we add columns 1 and 2, and compare the results to column 3). Firms with multiple relationships are less profitable, are larger, have a lower Tobin's Q and have a higher other intangibility and debt ratio. The latter result contrasts with the negative sign in most logit specifications and indicates the need to control for other relevant firm characteristics in an empirical model.

In Table V, we focus in more detail on the bilateral switching decision and its correspondence with profitability. We include in the sample only firms with a bilateral relationship and positive total interest payments. This reduces our sample to 800 observations. Higher interest payments over book value of debt (INTEREST) result in significantly more switching, but the marginal effect at the means is very small. In addition, bilateral switching corresponds to substantially lower profitability (the coefficient on dSINGLE{0}). However, given the limited number of switches we are able to observe, we should be cautious in our interpretation of these results.

IV. Conclusions

Our empirical work suggests a noticeable correspondence between the sales profitability of Norwegian publicly listed firms and their banking arrangements. Firms with a bilateral relationship are more profitable; firms that are more profitable more often maintain a bilateral relationship. This result is economically meaningful and robust. It holds under a variety of specifications controlling for firm age, size, debt, goodwill, other asset intangibility, and Tobin's Q.

The effects of the replacement, start-up, or termination of one of multiple relationships are less clear. Replacing one of multiple relationships and starting a new relationship seems to decrease profitability for two years or more following the event. However, these results are not robust to shortening in the length of the event window or to model alterations, possibly because the sample does not contain enough events. Switching firms are, on average, smaller and younger. They also seem to be less profitable and pay a higher interest rate in the switching year than other single bank firms across their listing period. Hence, mainly "bad" firms switch.

A life-cycle view of firm financing may synthesize the results. Small, young firms with promising projects start with bilateral relationships. If they grow and continue to have high quality projects, then these firms may remain with the same bank, thus strengthening their performance. If firms grow, but their future projects are only mediocre, then firms may switch or add more relationships in order to satisfy their financial needs, mitigate hold-up, and/or convey their lower quality to competitors by maintaining multiple relationships.

(1.) Berger, Saunders, Scalise, and Udell (1998), Boot (2000), and Ongena and Smith (2000a) review the literature.

For comments on previous drafts, we thank two anonymous referees, Abe de Jong, Doris Neuberger, Joseph Plasmans, Gordon Roberts, Kristian Rydqvist, Lemma W. Senbet, James K. Seward, David C. Smith, Elmer Sterken, Oved Yosha, and Rudi Vander Vennet. We also thank seminar participants at the 1999 European Economic Association Meetings (Santiago de Compostella). the 1999 Financial Management Association Meetings (Barcelona), the Norwegian School of Management BI, and the Universities of Gent, Groningen, Leuven, Rostock, and Tilburg (CentER). We thank Dog Michalsen for providing Norwegian bank data. We have benefited from the research assistance of Qinglei Dai and the editorial assistance of Jenny Bovenberg. The research was partly supported by the Fund for the Advancement of Bank Education at the Norwegian School of Management BI and the Foundation for Scientific Research Flanders under contract G.0302.00.

(*.) Hans Degryse is an Assistant Professor in Financial Economics at K.U.Leuven and Tilburg University. Steven Ongena is an Assistant Professor in Finance at CentER. Tilburg University in the Netherlands.

(2.) On the other hand, banks may still occasionally, deliberately, or accidentally, leak confidential information. The legal fallout for the bank of such an information spill may be limited; banks may try to optimally "liquefy reputation", and choose to provide useful information to an important client. Hellwig (1991), for example, notes that "the provision of information and advice to firms has also been an important part of the bank-firm relationship. Today this aspect may be even more important: if Deutsche Bank is involved with the acquisition program of Daimler-Benz, the relationship probably has more to do with Daimler-Benz's demand for information about other firms and industries than with the moral hazard problems in the relation between Daimler Benz and its financiers..." German banks are often suspected of shepherding domestic firms through their holdings of voting rights and board membership and resulting informational leverage. For example, on March 18th, 1997, Krupp launched a hostile takeover b id for Thyssen. Deutsche Bank and Dresdner Bank were important holders of voting rights at Thyssen. To back the hostile takeover bid, both "Deutsche Morgan Grenfell and Dresdner Kleinwort Bentwort, provided, through their parent commercial banks, a credit line believed to be worth DM18 billion" (The Economist, March 22, 1997). Deutsche was criticised because one of its board members was also a member of the Thyssen supervisory board. "He sat alongside the Thyssen management at the company's annual meeting last Friday without giving a hint of the bid plans or the bank's role" (Financial Times, March 21, 1997). "The deal faltered mainly because Deutsche Bank and Dresdner Bank ... got cold feet" (The Economist, March 29, 1997).

(3.) Other papers have touched indirectly upon the connection between the number of creditors and firm performance. For example, Horiuchi (1994) reports that statistical tests do not reveal any significant differences among Japanese firms having one, two, or three main banks as regards their profit to asset ratio. However, the firms in Horiuchi's sample use more banks than the reported "main" banks. No statistical tests are reported concerning the total number of bank relationships and the profit to asset ratio. Similarly, Houston and James (1996) find that the profitability of US publicly listed firms with one versus multiple bank relationships do not differ significantly. They focus on the reliance on bank debt, which they find negatively related to the importance of growth opportunities for firms with a single bank relationship and positively related for firms borrowing from multiple banks.

(4.) A Herfindahl-Hirschman Index, calculated as the sum of the squared percentage shares of the total number of relationships with publicly listed firms held by each bank, stood at 3258 in 1989, and 2984 in 1994. The US Department of Justice and Federal Trade Commission's Horizontal Merger Guidelines (April 1992) label markets with an HHI above 1800 as "highly concentrated." The 3-Bank and 5-Bank concentration ratios for Norway for the period 1989-1996 are 0.6 and 0.74 respectively, which are substantially higher than the comparable ratios for the UK and the US (Cetorelli and Gambera, 2000, Table I).

(5.) In 1995, more than 90% of the firms with multilateral relationships used both DnB and CBK (Norway's two largest banks).

(6.) We remove 14 observations for which the percentage ratio of earnings before interest and taxes to end-of-year sales is less than -150%, and two observations with negative sales values. Firms with extreme negative levels of profitability are likely to be atypical and the removal of extreme observations may facilitate the interpretation of the results.

(7.) Bhattacharya and Chiesa (1995), Yosha (1995), and von Rheinbaben and Ruckes (1998) model bilateral versus multilateral financing arrangements. Detragiache, Garella, and Guiso (2000) distinguish between an equilibrium with only one bank relationship and with multiple banking.

(8.) In a linear model, we cannot constrain the fitted dRELATION to the 0-1 interval. Greene (1997) lists other possible problems with the linear model (p. 874). We also re-estimate all specifications using a probit model, as "it is difficult to justify the choice of {logit or probit} on theoretical grounds" (p. 875). Results are virtually unaffected.

(9.) The industry classification scheme employed by Finansavisen assigns firms to eight different business categories: manufacturing, media, off-shore (i.e., mainly oil- and gas-related activities), shipping, other transportation, IT, real estate, and distribution. Our assignment of firms to categories on the basis of their 1995 classification is problematic for firms not listed in 1995 (which we assign to a "unclassified" category), for firms that shifted industry during the sample years, and for firms that operate in more than one industry.

(10.) The coefficients in the two-stage procedure are consistent, but the obtained standard errors are biased. To calculate unbiased standard errors in a linear-linear model, we multiply the obtained standard errors by the ratio of the standard error of the residuals from a second-stage regression using the actual instrumented variable (labeled 'A' in Table III) and the standard error of the residuals from a second-stage regression using the fitted instrumented variable (labeled 'F' in Table III). (See, for example, Gujarati, 1995, p. 705 for details). In a linear-logit model, this adjustment provides a reasonable first-order approximation for the unbiased standard errors.

(11.) Marginal effects vary with the values of the independent variables. (See, for example, Greene, 1997, p. 876).

(12.) Only four coefficients are positive, while seven coefficients are negative and significant at a 10% level.

(13.) This is only the case if the quality of innovators is constant over a longer period, in which case information possessed by the former bank remains accurate. On the other hand, if the quality of innovators changes frequently and information about their quality is quickly outdated, then innovators may still wish to replace their bank since banks have different market shares. In alternative extensions of Yosha's (1995) model, a replacement could result from coordination among incumbant and entrant (in order to avoid or propagate leakage of information). The quality of the signal could depend on the quality of the bank.

(14.) The interaction terms would be a perfect predictor of the dependent variable in such a specification (see, for example, Greene, 1997, p. 892).

(15.) We can reject the null of no overdispersion in the Poisson regression model using a regression-based test (Cameron and Trivedi, 1990), and turn to the negative binomial model, as it is the most commonly used alternative (Greene, 1997).

(16.) For the negative binomial model, we report standard deviations of the residuals from the corresponding linear model.

(17.) We actually discard the first year in each listing period because it is impossible to know whether switching occurred.

References

Angelini, P., R. Di Salvo, and G. Fern, 1998, "Availability and Cost of Credit for Small Businesses: Customer Relationships and Credit Cooperatives," Journal of Banking and Finance 22, 925-54.

Ber, H., Y. Yafeh, and O. Yosha, 2000, "Conflict of Interest in Universal Banking: Bank Lending, Stock Underwriting, and Fund Management," Journal of Monetary Economics (forthcoming).

Berger, A.N., A. Saunders, J.M. Scalise, and G.F. Udell, 1998, "The Effects of Bank Mergers and Acquisistions on Small Business Lending," Journal of Financial Economics 50, 187-230.

Bhattacharya, S. and G. Chiesa, 1995, "Proprietary Information, Financial Intermediation, and Research Incentives," Journal of Financial Intermediation 4, 328-57.

Bolton, P. and D.S. Scharfstein, 1996, "Optimal Debt Structure and the Number of Creditors," Journal of Political Economy 104, 1-25.

Boot, A.W.A., 2000, "Relationship Banking: What Do We Know?," Journal of Financial Intermediation 9, 3-25.

Boot, A.W.A. and A.V. Thakor, 1994, "Moral Hazard and Secured Lending in an Infinitely Repeated Credit Market Game," International Economic Review 35, 899-920.

Brander, J.A. and T.R. Lewis, 1986, "Oligopoly and Financial Structure: the Limited Liability Effect," American Economic Review 76, 956-70.

Cameron, A. and P. Trivedi, 1990, "Regression Based Tests for Overdispersion in the Poisson Model," Journal of Econometrics 31, 255-74.

Campbell, T.S., 1979, "Optimal Investment Financing Decisions and the Value of Confidentiality," Journal of Financial and Quantitative Analysis 14,232-57.

Cetorelli, N. and M. Gambera, 2000, "Banking Market Structure, Financial Dependence and Growth: International Evidence from Industry Data," Journal of Finance (forthcoming).

Cole, R., 1998, "The Importance of Relationships to the Availability of Credit," Journal of Banking and Finance 22, 959-77.

Degryse, H. and P. Van Cayseele, 2000, "Relationship Lending Within a Bank-Based System; Evidence from European Small Business Data," Journal of Financial Intermediation 9, 90-109.

Detragiache, E., P.G. Garella and L. Guiso, 2000, "Multiple versus Single Banking Relationships: Theory and Evidence," Journal of Finance 55, 1133-61.

Diamond, D., 1991, "Monitoring and Reputation: the Choice between Bank Loans and Privately Placed Debt," Journal of Political Economy 99, 689-721.

Drees, B. and C. Pazarbasioglu, 1995, "The Nordic Banking Crises: Pitfalls in Financial Liberalization," International Monetary Fund Working Paper 61.

Economist, The, 1997, "Exaggerated rumours of a death," (March 22) 8009,79-80.

Economist, The, 1997, "Auf Wiedersehen, shareholders," (March 29) 8010, 68.

Elsas, R. and J.P. Krahnen, 1998, "Is Relationship Lending Special? Evidence from Credit-File Data in Germany," Journal of Banking and Finance 22, 1283-316.

Fama, E.F. and K.R. French, 1997, "Dividends, Debt, Investment, and Earnings," University of Chicago mimeo.

Fama, E.F. and J.D. MacBeth, 1973, "Risk, Return, and Equilibrium: Empirical Tests," Journal of Political Economy 81, 607-36.

Financial Times, London Ed., 1997, "NEWS: EUROPE: Russian Capital flight surge EUROPEAN NEWS DIGEST:" (March 21) 02.

Finansavisen, 1995, Section Stockprices, (December 23).

Fischer, K., 1990, "Hausbankbeziehungen als Instrument der Bindung zwischen Banken und Unternehmen - Eine Theoretische und Empirische Analyse," Universitat Bonn PhD dissertation.

Foglia, A., S. Laviola, and P. Marullo Reedtz, 1998, "Multiple Banking Relationships and the Fragility of Corporate Borrowers," Journal of Banking and Finance 22, 1441-56.

Gorton, G. and F.A. Schmid, 2000, "Universal Banking and the Performance of German Firms," Journal of Financial Economics 58 (forthcoming).

Greene, W.H., 1995, Limdep 7.0 Manual, Econometric Software Inc.

Greene, W.H., 1997, Econometric Analysis, Upper Saddle River NJ, Prentice Hall.

Gujarati, D.N., 1995, Basic Econometrics, New York, McGraw-Hill International Editions.

Harhoff, D. and T. Korting, 1998a, "How Many Creditors Does it Take to Tango?," Wissenschaftszentrum Berlin mimeo.

Harhoff, D. and T. Korting, 1998b, "Lending Relationships in Germany - Empirical Evidence from Survey Data," Journal of Banking and Finance 22, 1317-53.

Harris, M. and A. Raviv, 1991, "The Theory of Capital Structure," Journal of Finance 46, 297-355.

Hellwig, M., 1991, "Banking, Financial Intermediation and Corporate Finance," in A. Giovannini and C.P. Mayer, Eds., European Financial Integration, Cambridge, Cambridge University Press, 35-63.

Horiuchi, T., 1994, "The Effect of Firm Status on Banking Relationships and Loan Syndication," in M. Aoki and H. Patrick, Eds., The Japanese Main Bank System, Oxford, Oxford University Press, 258-94.

Houston, J. and C. James, 1996, "Bank Information Monopolies and the Mix of Private and Public Debt Claims," Journal of Finance 51, 1863-89.

Hubert, F., 1998, "Financial Contracting when Rivals May Turn Nasty," Freie Universitat Berlin mimeo.

Jensen, M., 1986, "Agency Costs of Free Cash Flow, Corporate Finance and Takeovers," American Economic Review 76, 323-9.

Kang, J.K., A. Shivdasani, and T. Yamada, 2000, "The Effect of Bank Relations on Investment Decisions: An Investigation of Japanese Takeover Bids," Journal of Finance 55, 2197-2218.

Kierulf's Handbook, Oslo Boers Informasjon (Oslo Stock Exchange Information).

Machauer A. and M. Weber, 1999, "Number of Bank Relationships: An Indicator of Competition, Borrower Quality, or just Size," University of Mannheim mimeo.

Miarka, T., 1999, "On Bank-Firm Relationships and Bank-Debt: Evidence from Recent Developments in Japan," WZB mimeo.

Miarka, T., 2000, Financial Intermediation and Deregulation: A Critical Analysis of Japanese Bank-Firm Relationships, Heidelberg, Springer Verlag.

Modigliani, F. and M.H. Miller, 1963, "The Cost of Capital, Corporation Finance and the Theory of Investment," American Economic Review 53, 26 1-97.

Myers, S. and N.S. Majluf, 1984, "Corporate Financing and Investment Decisions when Firms Have Information that Investors Do not Have," Journal of Financial Economics 13, 187-221.

Ongena, S. and D.C. Smith, 2000a, "Bank Relationships: A Survey," in P. Harker and S.A. Zenios, Eds., The Performance of Financial Institutions, Cambridge, Cambridge University Press.

Ongena, S. and D.C. Smith, 2000b, "The Duration of Bank Relationships," Journal of Financial Economics (forthcoming).

Ongena, S. and D.C. Smith, 2000c, "What Determines the Number of Bank Relationships? Cross-Country Evidence," Journal of Financial Intermediation 9, 26-56.

Petersen, M.A. and R.G. Rajan, 1994, "The Benefits of Lending Relationships: Evidence from Small Business Data," Journal of Finance 49, 3-37.

Petersen, M.A. and R.G. Rajan, 1995, "The Effect of Credit Market Competition on Lending Relationships," Quarterly Journal of Economics 110, 406-43.

Rajan, R.G., 1992, "Insiders and Outsiders: the Choice between Informed and Arm's-length Debt," Journal of Finance 47, 1367-400.

Ross, S.A., 1977, "The Determination of Financial Structure: the Incentive-Signaling Approach," Bell Journal of Economics 8, 1-30.

Scott, J.A., 2000, "Relationships, Access to Credit and Loan Pricing: an Analysis of Small Business Experience," Temple University mimeo.

Sharpe, S.A., 1990, "Asymmetric Information, Bank Lending and Implicit Contracts: a Stylized Model of Customer Relationships," Journal of Finance 45, 1069-87.

Stancill, J.M., 1980, "Getting the Most from Your Banking Relationship," Harvard Business Review 80, 141-5.

Statistical Yearbook of Norway, 1996, Statistisk SentralByraa (Statistics Norway).

Stulz, R.M., 1990, "Managerial Discretion and Optimal Financing Policies," Journal of Financial Economics 19, 3-27.

Titman, S., 1984, "The Effect of Capital Structure on a Firm's Liquidation Decision.," Journal of Financial Economics 13, 137-51.

US Department of Justice and the Federal Trade Commission, 1997, Horizontal Merger Guidelines, (April 8), [less than]http://www.usdoj.gov/atr/public/guidelines/horiz_book/hmg1.html[ greater than]

von Rheinbaben, J. and M. Ruckes, 1998, "A Finn's Optimal Number of Bank Relationships and the Extent of Information Disclosure," University of Mannheim mimeo.

von Thadden, E.L., 1992, "The Commitment of Finance, Duplicated monitoring and the Investment Horizon," ESF-CEPR Working Paper in Financial Markets 27.

von Thadden, E.L., 1998, "Asymmetric Information, Bank Lending, and Implicit Contracts: the Winner's Curse," University of Lausanne (DEEP) mimeo.

Weinstein, D.E. and Y. Yafeh, 1998, "On the Costs of a Bank Centered Financial System: Evidence from the Changing Main Bank Relations in Japan," Journal of Finance 53, 635-72.

Yosha, O., 1995, "Information Disclosure Costs and the Choice of Financing Source," Journal of Financial Intermediation 4, 3-20.

Full Sample Characteristics

Firms report a minimum of one and a maximum of four "primary" bank relationships listed in Kierulf's Handbook. The total number of firms reporting relationships is the number of different firms listing relationships in Kierulf's Handbook between 1979 and 1995. The total number of firm years is the number of firms listing a relationship times the number of years the firm has reported the relationship in Kierulf's Handbook between 1979 and 1995. The total number of relationship years is the number of firms listing relationships times the number of years the firm has reported the relationship times the number of events (or is one, if no events take place) for that particular firm. Events include the replacement of a single relationship, the replacement of one of multiple relationships, the startup of an additional relationship, and the end of a relationship. The data from Kierulf's Handbook is matched with accounting information gleaned from FINLIS, an accounting database supported by Oslo Bors Informasjon, a su bsidiary of the Oslo Stock Exchange.
 Kierulf's After matching
 Handbook with FINLIS
Total number of firms reporting bank
relationships 235 176
Total number of firm years 1897 1284
Total number of relationship years 2436 1659
Total number of events 118 87
Replacement of a single relationship 44 30
Replacement of one of multiple
relationships 16 15
Start-up of an additional relationship 24 18
End of a relationship 34 24


Descriptive Statistics of Firm Characteristics

The number of firm-year observations is 1284, except for the ownership concentration ratio which is 883. Gross ROA is the percentage ratio of earnings before interest and taxes to the sum of the market value of equity, plus the book value of debt. Pre-tax income/assets is the percentage ratio of earnings before taxes to the sum of the market value of equity, plus the book value of debt. EBIT/sales is the percentage ratio of earnings before interest and taxes to end-of-year sales. Sales is the end-of-year sales, deflated by the Norwegian CPI (1979=1). Assets is the sum of the market value of equity, plus the book value of debt, deflated by the Norwegian CPI (1979=1). The debt ratio is the ratio of the book value of debt to the sum of the market value of equity, plus the book value of debt. The age of the firm is measured relative to the founding date of the firm. Goodwill (other intangibles) is the ratio of the total book value of goodwill (or other intangibles) to the sum of the market value of equity, plus t he book value of debt. Tobin's Q is the ratio of the end-of-year market value of equity, plus the book value of debt, divided by the book value of assets. The ownership concentration ratio is the proportion of firm equity owned by the ten largest shareholders. The multiple-bank relationship dummy takes the value of one when a firm maintains a multiple-relationship. It is zero otherwise. Maximum market share of relationship bank(s) is the (maximum) proportion of all registered bank connections accounted for by the firm's bank(s). Total interest payments are divided by the book value of debt. A t-test determines the statistical significance of the correlation coefficients (number of observations=883 for correlation coefficients with the ownership concentration ratio; otherwise =1284).
Variable Mean St. Dev. Minimum
Gross ROA (in %) 3.23 9.50 -57.38
Pre-tax income / assets (in %) 0.12 10.33 -86.09
EBIT / sales (in %) 6.38 21.24 -141.73
Sales (in thousands of kroner) [a] 900,616 2,736,829 4,777
Number of employees 1780 4138 1
Assets (in thousands of kroner) [a] 1,189,793 3,513,813 4,782
Debt ratio 0.614 0.223 0
Age 59 41 0
Goodwill 0.006 0.019 0
Other intangibles 0.006 0.027 0
Tobin's Q 1.401 1.017 0.385
Ownership concentration ratio 0.660 0.206 0
Multiple-bank relationship dummy 0.290 0.454 0
Maximum market share of relationship 0.296 0.121 0.005
bank(s)
DnB and/or CBK as main bank 0.895 0.305 0
Total interest payments / debt (in %) 5.02 4.45 0.00
Variable 25% Median 75%
Gross ROA (in %) 1.44 5.03 7.72
Pre-tax income / assets (in %) -1.85 2.16 4.52
EBIT / sales (in %) 1.69 6.00 10.99
Sales (in thousands of kroner) [a] 645,380 1,789,300 6,304,600
Number of employees 116 486 1461
Assets (in thousands of kroner) [a] 112,880 262,750 835,130
Debt ratio 0.453 0.639 0.790
Age 19 63 86
Goodwill 0 0 0.001
Other intangibles 0 0 0.001
Tobin's Q 1.038 1.202 1.504
Ownership concentration ratio 0.525 0.674 0.810
Multiple-bank relationship dummy 0 0 1
Maximum market share of relationship 0.267 0.293 0.370
bank(s)
DnB and/or CBK as main bank 1 1 1
Total interest payments / debt (in %) 0.00 5.31 7.68
Variable Maximum
Gross ROA (in %) 23.1
Pre-tax income / assets (in %) 20.42
EBIT / sales (in %) 85.7
Sales (in thousands of kroner) [a] 29,864,000
Number of employees 43122
Assets (in thousands of kroner) [a] 46,340,000
Debt ratio 0.997
Age 245
Goodwill 0.181
Other intangibles 0.473
Tobin's Q 30.730
Ownership concentration ratio 1
Multiple-bank relationship dummy 1
Maximum market share of relationship 0.481
bank(s)
DnB and/or CBK as main bank 1
Total interest payments / debt (in %) 57.19
(a.)8 kroners bought approximately 1 US $ in March 2000.
 Descriptive Statistics of Firm
 Characteristics
Correlation Coefficients (2) (3)
(1) Gross ROA 0.957 [***] 0.562 [***]
(2) Pre-tax income/assets 1 0.541 [***]
(3) EBIT/sales 1
(4) Sales
(5) Number of employees
(6) Assets
(7) Debt ratio
(8) Age
(9) Goodwill
(10) Other Intangibles
(11) Tobin's Q
(12) Ownership concentration ratio
(13) Multiple-bank relationship dummy
(14) Market share of relationship bank(s)
(15) DnB and/or CBK as a main bank
(16) Total interest payments/ debt (in %)
Correlation Coefficients (4) (5)
(1) Gross ROA 0.100 [***] 0.093 [***]
(2) Pre-tax income/assets 0.095 [***] 0.091 [***]
(3) EBIT/sales 0.018 -0.004
(4) Sales 1 0.940 [***]
(5) Number of employees 1
(6) Assets
(7) Debt ratio
(8) Age
(9) Goodwill
(10) Other Intangibles
(11) Tobin's Q
(12) Ownership concentration ratio
(13) Multiple-bank relationship dummy
(14) Market share of relationship bank(s)
(15) DnB and/or CBK as a main bank
(16) Total interest payments/ debt (in %)
Correlation Coefficients (6) (7)
(1) Gross ROA 0.205 [***] -0.158 [***]
(2) Pre-tax income/assets 0.195 [***] -0.264 [***]
(3) EBIT/sales 0.233 [***] -0.166 [***]
(4) Sales 0.551 [***] 0.067 [**]
(5) Number of employees 0.630 [***] 0.116 [***]
(6) Assets 1 0.049 [*]
(7) Debt ratio 1
(8) Age
(9) Goodwill
(10) Other Intangibles
(11) Tobin's Q
(12) Ownership concentration ratio
(13) Multiple-bank relationship dummy
(14) Market share of relationship bank(s)
(15) DnB and/or CBK as a main bank
(16) Total interest payments/ debt (in %)
Correlation Coefficients (8) (9)
(1) Gross ROA 0.039 0.089 [***]
(2) Pre-tax income/assets -0.268 [***] 0.099 [***]
(3) EBIT/sales -0.055 [**] 0.026
(4) Sales 0.106 [***] 0.019
(5) Number of employees 0.145 [***] 0.042
(6) Assets 0.016 0.095 [***]
(7) Debt ratio 1.140 [***] -0.096 [***]
(8) Age 1 -0.083 [***]
(9) Goodwill 1
(10) Other Intangibles
(11) Tobin's Q
(12) Ownership concentration ratio
(13) Multiple-bank relationship dummy
(14) Market share of relationship bank(s)
(15) DnB and/or CBK as a main bank
(16) Total interest payments/ debt (in %)
Correlation Coefficients (10) (11)
(1) Gross ROA 0.024 0.071 [**]
(2) Pre-tax income/assets 0.020 0.118 [***]
(3) EBIT/sales -0.000 0.175 [***]
(4) Sales 0.064 [**] -0.035
(5) Number of employees 0.038 -0.056 [**]
(6) Assets 0.101 [***] 0.018
(7) Debt ratio -0.021 -0.482 [***]
(8) Age -0.098 [***] -0.062 [**]
(9) Goodwill -0.029 0.004
(10) Other Intangibles 1 -0.022
(11) Tobin's Q 1
(12) Ownership concentration ratio
(13) Multiple-bank relationship dummy
(14) Market share of relationship bank(s)
(15) DnB and/or CBK as a main bank
(16) Total interest payments/ debt (in %)
Correlation Coefficients (12) (13)
(1) Gross ROA -0.159 [***] -0.128 [***]
(2) Pre-tax income/assets -0.178 [***] -0.155 [***]
(3) EBIT/sales -0.023 -0.076 [***]
(4) Sales -0.140 [***] 0.075 [***]
(5) Number of employees -0.131 [***] 0.148 [***]
(6) Assets -0.081 [**] 0.317 [***]
(7) Debt ratio 0.036 0.248 [***]
(8) Age -0.010 0.062 [**]
(9) Goodwill 0.020 -0.016
(10) Other Intangibles -0.154 [***] 0.069 [**]
(11) Tobin's Q -0.034 -0.122 [***]
(12) Ownership concentration ratio 1 0.018
(13) Multiple-bank relationship dummy 1
(14) Market share of relationship bank(s)
(15) DnB and/or CBK as a main bank
(16) Total interest payments/ debt (in %)
Correlation Coefficients (14) (15)
(1) Gross ROA -0.060 [**] 0.091 [***]
(2) Pre-tax income/assets -0.063 [***] -0.012
(3) EBIT/sales 0.015 0.050 [*]
(4) Sales 0.137 [***] 0.095 [***]
(5) Number of employees 0.166 [***] 0.119 [***]
(6) Assets 0.299 [***] 0.224 [***]
(7) Debt ratio 0.048 [*] 0.095 [***]
(8) Age -0.004 0.008
(9) Goodwill 0.030 -0.030
(10) Other Intangibles 0.011 0.056 [**]
(11) Tobin's Q -0.074 [**] -0.083 [***]
(12) Ownership concentration ratio 0.044 -0.025
(13) Multiple-bank relationship dummy 0.299 [***] 0.224 [***]
(14) Market share of relationship bank(s) 1 0.735 [***]
(15) DnB and/or CBK as a main bank 1
(16) Total interest payments/ debt (in %)
Correlation Coefficients (16)
(1) Gross ROA -0.095 [***]
(2) Pre-tax income/assets -0.326 [***]
(3) EBIT/sales -0.056 [**]
(4) Sales -0.029
(5) Number of employees -0.055 [**]
(6) Assets -0.047 [*]
(7) Debt ratio 0.028
(8) Age 0.061 [**]
(9) Goodwill -0.029
(10) Other Intangibles 0.018
(11) Tobin's Q -0.039
(12) Ownership concentration ratio 0.096 [***]
(13) Multiple-bank relationship dummy -0.047 [*]
(14) Market share of relationship bank(s) -0.000
(15) DnB and/or CBK as a main bank 0.051 [*]
(16) Total interest payments/ debt (in %) 1
(***.)Significant at the 0.01 level.
(**.)Significant at the 0.05 level.
(*.)Significant at the 0.10 level.


OLS and 2SLS Specification

The Static Sample contains 1284 firm years; the Dynamic Sample contains 1659 relationship years. ROA is the percentage ratio of earnings before interest and taxes to the sum of the market value of equity, plus the book value of debt, i.e., gross return on assets. SALES is the log of the end-of-year sales (in thousands of kroner), deflated by 100 times the Norwegian CPI (1979=1). DEBT is the ratio of the book value of debt to the sum of the market value of equity, plus the book value of debt. AGE is the log of the age of the firm, and is measured relative to the founding date of the firm. GOODWILL (OTHER INTANGIBLES) is the ratio of the total book value of goodwill (or other intangibles) to the sum of the market value of equity, plus the book value of debt. Q is the ratio of the year-end market value of equity, plus book value of debt, divided by the book value of assets. dRELATION takes the value of one when a firm maintains a multiple-bank relationship, and is zero otherwise. RELATION equals the number of ba nk relationships a firm maintains (minimum 1, maximum 4). dSINGLE {0,l,2} takes the value of one in the first year and up to two years after a firm replaces a single relationship, and zero otherwise. dREPLACE (dSTART, dEND) {0,l,2} takes the value of one in the first year and up to two years after a firm replaces (starts, ends) one of multiple relationships, and zero otherwise. BANKSHARE is the (maximum) proportion of all registered bank connections accounted for by the firm's bank(s). dDnBorCBK, takes the value of one if the firm has a relationship with DnB (Bergens Bank or DnC before their merger) and/or CBK, and zero otherwise. Coefficients are listed on the first row in each cell. Heteroskedasticity-consistent standard errors are listed below in parentheses. Coefficients for instrumental variables are in bold-faced italics. Measures of goodness-of-fit are the adjusted-[R.sup.2] for the linear models, and the likelihood-ratio index for the logit models. Standard deviation of residuals is also calculated fo r a specification including the actual (A) and the fitted (F) values of the instrumented variable.
Model (1) (2)
Equation (a) (b)
Sample/Number of Obs. Static/1284 Dynamic/1659
 2S 2S
Estimation OLS Linear Logit
 Industry Industry
Effects No Year Year
Dependent Variable ROA ROA dRELATION
Constant -5.67 [***] 0.85 -5.43 [***]
 (1.50) (1.78) (0.82)
SALES 1.64 [***] 1.86 [***] 0.46 [***]
 (0.15) (0.13) (0.04)
DEBT -7.39 [***] -9.90 [***] 1.49 [***]
 (1.29) (1.44) (0.52)
AGE 0.46 [**] -0.12 -0.31 [***]
 (0.22) (0.21) (0.06)
GOODWILL 19.31 [**] 39.49 [***] -4.87
 (8.25) (8.19) (4.29)
OTHER INTANGIBLES 8.62 3.46 2.08
 (5.66) (4.56) (2.06)
Q 0.09 0.01 -0.24
 (0.14) (0.08) (0.20)
ROA -0.06 [***]
 (0.01)
dRELATION or -3.87 [**] -1.63 [***]
RELATION (0.72) (0.60)
Model (3)
Equation (a) (b)
Sample/Number of Obs. Dynamic/1659
 2S 2S
Estimation Linear Neg. Bin.
 Industry Industry
Effects Year Year
Dependent Variable ROA RELATION
Constant 3.18 -0.23
 (2.59) (0.47)
SALES 2.18 [***] 0.21 [***]
 (0.19) (0.05)
DEBT -5.96 [***] -0.19
 (1.63) (0.42)
AGE 0.03 -0.03
 (0.48) (0.07)
GOODWILL 30.24 -0.37
 (25.72) (11.27)
OTHER INTANGIBLES 6.13 -2.52
 (19.53) (4.22)
Q 0.06 0.009
 (0.08) (0.147)
ROA -0.08 [***]
 (0.03)
dRELATION or -4.16 [***]
RELATION (1.35)
dSINGLE{0,1,2} 1.46 [*]
 (0.86)
dREPLACE{0,1,2} -6.28 [**]
 (3.12)
dSTART{0,1,2} -1.85 [**]
 (0.89)
dEND{0,1,2} 0.94
 (0.81)
BANKSHARE 8.10 [***]
 (1.31)
dDnBorCBK -0.25
 (0.59)
GOODWILL [*] RELATION
OTHER INTANGIBLES [*]
RELATION
AGE [*] RELATION
Goodness of Fit 0.114 0.166 0.252
St. Dev. of Residuals (A) - 8.345 0.467
St. Dev. of Residuals (F) 8.942 8.436 0.473
Log Likelihood -4712.896 -5874.863 -800.019
dSINGLE{0,1,2} 0.60
 (0.90)
dREPLACE{0,1,2} -2.20
 (3.03)
dSTART{0,1,2} 0.82
 (1.26)
dEND{0,1,2} 0.40
 (0.83)
BANKSHARE 1.06
 (0.97)
dDnBorCBK -0.20
 (0.37)
GOODWILL [*] RELATION -0.66 1.59
 (19.84) (7.03)
OTHER INTANGIBLES [*] 5.74 1.55
RELATION (6.55) (1.19)
AGE [*] RELATION -0.014 [**] -0.0012
 (0.005) (0.011)
Goodness of Fit 0.163 0.059
St. Dev. of Residuals (A) 8.454 0.718
St. Dev. of Residuals (F) 8.420 0.684
Log Likelihood -6042.698 -2070.862
(***.)Significant at the 0.01 level.
(**.)Significant at the 0.05 level.
(*.)Significant at the 0.10 level.


Replacement of Single Bank Relationships

The first column lists the characteristics of firms replacing a single bank relationship in the event year. The second column reports characteristics of all other firms. Across their listing period they have only one relationship and do not replace that relationship. The last column lists characteristics of firms that report multiple relationships. Gross ROA is the percentage ratio of earnings before interest and taxes to the sum of the market value of equity, plus the book value of debt. Pre-tax income/assets is the percentage ratio of earnings before taxes to the sum of the market value of equity, plus the book value of debt. EBIT/sales is the percentage ratio of earnings before interest and taxes to end-of-year sales. Sales is the end-of-year sales, deflated by the Norwegian CPI (1979=1). Assets is the sum of the market value of equity, plus the book value of debt, deflated by the Norwegian CPI(1979=1). The debt ratio is the ratio of the book value of debt to the sum of the market value of equity, plus the book value of debt. The age of the firm is measured relative to the founding date of the firm. Goodwill (other intangibles) is the ratio of the total book value of goodwill (other intangibles) to the sum of the market value of equity, plus the book value of debt. Tobin's Q is the ratio of the year-end market value of equity plus book value of debt, divided by the book value of assets. Means are listed on the first row in each cell; standard errors are reported below in parentheses. Medians are listed on the third row. The means are compared using a two-sided T-test (variances are not assumed equal). To compare medians, we use a two-sided Mann-Whitney test.
 Bilateral Bilateral
Variable Replacing Not Replacing
Number of observations 30 538
Gross ROA (in %) 1.91 4.23
 (8.59) (7.75)
 4.32 5.39
Pre-tax income / assets (in %) -0.46 0.91
 (9.28) (8.14)
 2.12 2.41
EBIT / sales (in %) -1.32 9.10
 (34.41) (18.54)
 6.33 6.87
Sales (in thousands of kroner) 186,516 [***] 1,058,400
 (191,602) (4,031,300)
 121,420 130,740
Number of employees 387 [***] 1772
 (439) (5423)
 207 [***] 372
Assets (in thousands of kroner) 329,994 [***] 1,473,228
 (474,062) (5,117,825)
 177,820 173,520
Debt ratio 0.506 0.575
 (0.238) (0.225)
 0.572 0.594
Variable Multiple
Number of observations 574
Gross ROA (in %) 1.92 [a]
 (11.77)
 4.80 [a]
Pre-tax income / assets (in %) -1.95 [a]
 (13.47)
 0.93 [a]
EBIT / sales (in %) 5.00 [a]
 (17.69)
 5.75 [a]
Sales (in thousands of kroner) 1,208,700
 (1,539,000)
 580,220 [a]
Number of employees 2697 [a]
 (3612)
 1214 [a]
Assets (in thousands of kroner) 1,627,487
 (1,998,165)
 632,090 [a]
Debt ratio 0.707 [a]
 (0.184)
 0.724 [a]
Age 46 [**] 59 59
 (33) (42) (41)
 33 61 67
Goodwill 0.013 0.005 0.005
 (0.034) (0.020) (0.011)
 0 0 0
Other intangibles 0.016 0.015 0.010 [b]
 (0.035) (0.048) (0.036)
 0 0 0
Tobin's Q 1.466 1.454 1.203 [a]
 (0.659) (0.663) (0.356)
 1.252 1.237 1.121 [a]
Interest payments / Debt 0.070 0.056 0.054
 (0.119) (0.039) (0.048)
 0.035 0.059 0.055
(***.)Significantly different from 'Not Replacing' at the 0.01 level.
(**.)Significantly different from 'Not Replacing' at the 0.05 level.
(*.)Significantly different from 'Not Replacing' at the 0.10 level.
(a.)Significantly different from 'Bilateral'
(Columns 1 and 2) at the 0.01 level.
(b.)Significantly different from 'Bilateral'
(Columns 1 and 2) at the 0.05 level.
(c.)Significantly different from 'Bilateral'
(Columns 1 and 2) at the 0.10 level.


Von Thadden Specification

The Restricted Sample contains only relationship years for firms maintaining a bilateral bank relationship and reporting positive interest payments. ROA is the percentage ratio of earnings before interest and taxes to the sum of the market value of equity, plus the book value of debt (i.e., gross return on assets). SALES is the log of the end-of-year sales (in thousands of kroner), deflated by 100 times the Norwegian CPI (1979=1). DEBT is the ratio of the book value of debt to the sum of the market value of equity, plus the book value of debt. AGE is the log of the age of the firm and is measured relative to the founding date of the firm. GOODWILL (OTHER INTANGIBLES) is the ratio of the total book value of goodwill (other intangibles) to the sum of the market value of equity, plus the book value of debt. Q is the ratio of the year-end market value of equity, plus book value of debt divided by the book value of assets. dSINGLE{0} takes the value of one in the first year a firm replaces a single relationship, a nd zero otherwise. INTEREST is defined as the interest payments divided by the book value of debt. Coefficients are listed on the first row in each cell. Heteroskedasticity-consistent standard errors are listed below in parentheses. Coefficients for instrumental variables are in bold-faced italics. Measures of goodness-of-fit are the adjusted-[R.sup.2] for the linear models, and the likelihood-ratio index for the negative binomial or logit models.
Equation (a) (b)
Sample Restricted
Number of Obs. 800
Estimation 2S Linear 2S Logit
Effects Industry Year Industry Year
Dependent Variable ROA dSINGLE{0}
Constant 0.07 -3.81 [***]
 (2.42) (1.42)
SALES 1.50 [***]
 (0.14)
DEBT -6.72 [***]
 (1.86)
AGE 1.03 [***]
 (0.31)
GOODWILL 25.08 [***]
 (8.75)
OTHER INTANGIBLES 2.90
 (9.24)
Q -1.11 [**]
 (0.56)
ROA -0.05
 (0.10)
dSINGLE{0} -25.77 [**]
 (10.99)
INTEREST 0.14 [**]
 (0.06)
Goodness of Fit 0.250 0.362
Log Likelihood -2704.550 -54.874
(***.)Significant at the 0.01 level.
(**.)Significant at the 0.05 level.
(*.)Significant at the 0.10 level.


Appendix

Sample and variable definitions can be found in the notes to Table III. PRETAX is the percentage ratio of earnings before taxes to the sum of the market value of equity, plus the book value of debt. PROFITS is the ratio of earnings before interest and taxes to sales. OWNERSHIP CONCENTRATION measures the proportion of equity owned by a firm's ten largest shareholders. BANKSIZE is the ratio of the (maximum) asset size of the firm's bank(s) (in thousands of kroner) and the firm's assets (in kroner). Coefficients are listed on the first row in each cell. Heteroskedasticity-consistent standard errors are listed below in parentheses (in Panel B, Model 1 we list the sample standard deviations divided by the square root of the number of observations). Coefficients for instrumented variables are in bolded italics.
Panel A.
Model / Equation (1.a) (1.b)
Sample / Number of Observations Dynamic / 1659
Impact Period {0}
Estimation 2S Linear 2S Logit
Dependent Variable ROA dRELATION
Constant, SALES, DEBT, AGE,
GOODWILL, OTHER INTANGIBLES, included included
Q, Industry & Year Effects
BANKSHARE and dDnBorCBK included
ROA -0.06 [***]
 (0.01)
dRELATION -3.76 [***]
 (0.74)
dSINGLE{Impact Period) -0.42
 (1.43)
dREPLACE(Impact Period) -4.12
 (4.80)
dSTART{Irnpact Period} -1.01
 (1.51)
dEND{Impact Period) 2.05
 (1.37)
Goodness of Fit 0.168 0.250
Log Likelihood -5872.591 -802.224
Panel A.
Model / Equation (2.a) (2.b)
Sample / Number of Observations Dynamic / 1659 Dynamic / 1659
Impact Period {0, 1}
Estimation 2S Linear 2S Logit
Dependent Variable ROA dRELATION
Constant, SALES, DEBT, AGE,
GOODWILL, OTHER INTANGIBLES, included included
Q, Industry & Year Effects
BANKSHARE and dDnBorCBK included
ROA -0.05 [***]
 (0.01)
dRELATION -1.60 [***]
 (0.60)
dSINGLE{Impact Period) 0.97
 (0.97)
dREPLACE(Impact Period) -5.89
 (3.73)
dSTART{Irnpact Period} -1.46
 (0.96)
dEND{Impact Period) 0.36
 (0.97)
Goodness of Fit 0.159 0.250
Log Likelihood -5881.916 -801.397
Panel A.
Model / Equation (3.a) (3.b)
Sample / Number of Observations Dynamic / 1659
Impact Period {0, 1, 2}
Estimation 2S Linear 2S Logit
Dependent Variable ROA dRELATION
Constant, SALES, DEBT, AGE,
GOODWILL, OTHER INTANGIBLES, included included
Q, Industry & Year Effects
BANKSHARE and dDnBorCBK included
ROA -0.06 [***]
 (0.01)
dRELATION -1.63 [***]
 (0.60)
dSINGLE{Impact Period) 1.46 [*]
 (0.86)
dREPLACE(Impact Period) -6.28 [**]
 (3.12)
dSTART{Irnpact Period} -1.85 [**]
 (0.89)
dEND{Impact Period) 0.94
 (0.81)
Goodness of Fit 0.166 0.252
Log Likelihood -6042.698 -800.019
Panel A.
Model / Equation (4.a) (4.b)
Sample / Number of Observations Dynamic / 1659
Impact Period {0, 1, 2, 3}
Estimation 2S Linear 2S Logit
Dependent Variable ROA dRELATION
Constant, SALES, DEBT, AGE,
GOODWILL, OTHER INTANGIBLES, included included
Q, Industry & Year Effects
BANKSHARE and dDnBorCBK included
ROA -0.06 [***]
 (0.01)
dRELATION -1.71 [***]
 (0.60)
dSINGLE{Impact Period) 1.01
 (0.78)
dREPLACE(Impact Period) -7.14 [**]
 (2.82)
dSTART{Irnpact Period} -1.76 [**]
 (0.80)
dEND{Impact Period) 0.64
 (0.73)
Goodness of Fit 0.173 0.252
Log Likelihood -5868.238 -799.547
Panel B.
Model / Equation (1.a) (1.b) (2.a)
Sample / Number of
Observations F-MacB / 15 Dynamic / 1659
Estimation OLS OLS 2S Linear
 Industry,
Effects No No Year
Dependent Variable ROA dRELATlON PRETAX
Constant -6.19 [**] -13.58 [***] 1.56
 (2.66) (1.29) (1.94)
SALES 1.94 [***] 0.79 [***] 1.83 [***]
 (0.19) (0.10) (0.15)
DEBT -10.32 [***] 0.79 -15.23 [***]
 (2.67) (0.78) (1.54)
AGE -0.02 -0.16 [***] -0.26
 (0.17) (0.04) (0.22)
GOODWILL 26.13 [*] -40.14 [***] 39.86 [***]
 (18.62) (33.91) (8.50)
OTHER 29.99 [**] 17.29 [*] 0.03
INTANGIBLES (16.99) (12.19) (4.81)
Q 0.53 -0.95 [***] 0.01
 (0.54) (0.32) (0.08)
SALES{-1}
OWNERSHIP
CONCENTRATION
ROA, PRETAX or -0.05 [***]
PROFITS (0.01)
Panel B.
Model / Equation (2.b) (3.a) (3.b)
Sample / Number of
Observations Dynamic / 1659
Estimation 2S Logit 2S Linear 25 Logit
 Industry, Industry, Industry,
Effects Year Year Year
Dependent Variable dRELATION PROFITS dRELATION
Constant -1.68 [*] -1.54 -6.82 [***]
 (0.98) (5.89) (0.87)
SALES 2.82 [***] 2.71 [***] 1.81 [***]
 (0.40) (0.52) (0.17)
DEBT -19.13 [***] -5.06 -2.18 [***]
 (3.42) (4.21) (0.71)
AGE -0.63 -2.52 [***] -1.69 [***]
 (0.08) (0.63) (0.17)
GOODWILL 44.21 [***] 10.78 2.06
 (9.23) (15.31) (4.62)
OTHER 1.53 11.89 8.31 [***]
INTANGIBLES (2.15) (24.53) (2.28)
Q -0.28 2.67 [***] 1.53 [***]
 (0.22) (0.72) (0.30)
SALES{-1}
OWNERSHIP
CONCENTRATION
ROA, PRETAX or -1.37 [***] -0.64 [***]
PROFITS (0.22) (0.07)
Panel B.
Model / Equation (4.a) (4.b) (5.a)
Sample / Number of Limited / 877 Limited/ 1047
Observations Limited / 877
Estimation 2S Linear 2S Logit 2S Linear
 Industry,
Effects Industry Industry Year
Dependent Variable ROA dRELATION ROA
Constant -3.40 -4.32 [***] 0.53
 (2.08) (0.95) (3.34)
SALES 2.09 [***] 0.67 [***] 2.46 [***]
 (0.20) (0.07) (0.26)
DEBT -12.02 [***] 0.86 -9.50 [***]
 (2.24) (0.71) (1.96)
AGE 0.10 -0.18 [**] -0.54 [**]
 (0.26) (0.07) (0.27)
GOODWILL 41.00 [***] -6.49 30.29 [***]
 (10.51) (5.14) (9.16)
OTHER 9.83 6.31 [**] 0.17
INTANGIBLES (6.65) (2.69) (8.35)
Q 0.005 -0.71 [**] -0.07
 (0.079) (0.30) (0.65)
SALES{-1} 0.31 [***]
 (0.08)
OWNERSHIP -2.38 [**]
CONCENTRATION (1.07)
ROA, PRETAX or -0.04 [***]
PROFITS (0.01)
Panel B.
Model / Equation (5.b)
Sample / Number of
Observations Limited/ 1047
Estimation 2S Logit
 Industry,
Effects Year
Dependent Variable dRELATION
Constant -7.95 [***]
 (1.36)
SALES 2.32 [***]
 (0.30)
DEBT -4.91 [**]
 (1.90)
AGE -0.45 [***]
 (0.09)
GOODWILL 18.25 [**]
 (8.70)
OTHER 4.49 [*]
INTANGIBLES (4.24)
Q -0.13
 (0.37)
SALES{-1} 0.37 [**]
 (0.17)
OWNERSHIP 1.93 [***]
CONCENTRATION (0.58)
ROA, PRETAX or -0.91 [***]
PROFITS (0.16)
Panel B
Model/Equation (1.a) (1.b) (2.a)
Sample/Number of
Observations F-MacB/15 Dynamic/1659
Estimation OLS OLS 2S Linear
 Industry,
Effects No No Year
Dependent Variable ROA dRELATION PRETAX
dRELATION or -3.89 [***] -0.72
RELATION (1.60) (0.67)
dSINGLE{0,1,2} 2.16 [**]
 (0.96)
dREPLACE{0,1,2} -6.41 [**]
 (3.26)
dSTART{0,1,2} -2.14 [**]
 (0.93)
dEND{0,1,2} 0.59
 (0.96)
BANKSHARE
dDnBorCBK
BANKSIZE
Goodness of Fit - - 0.175
Log Likelihood - - -6214.958
Panel B
Model/Equation (2.b) (3.a) (3.b)
Sample/Number of
Observations Dynamic/1659
Estimation 2S Logit 2S Linear 2S Logit
 Industry, Industry, Industry,
Effects Year Year Year
Dependent Variable dRELATION PROFITS dRELATION
dRELATION or -3.99 [***]
RELATION (1.49)
dSINGLE{0,1,2} 3.52
 (3.09)
dREPLACE{0,1,2} -1.93
 (2.93)
dSTART{0,1,2} -3.62
 (2.53)
dEND{0,1,2} -2.72
 (3.20)
BANKSHARE -5.45 [*] -2.61
 (2.78) (1.87)
dDnBorCBK 2.21 [***] 4.48 [***]
 (0.78) (0.84)
BANKSIZE
Goodness of Fit 0.306 0.112 0.298
Log Likelihood -741.701 -7459.113 -750.326
Panel B
Model/Equation (4.a) (4.b) (5.a)
Sample/Number of
Observations Limited/877 Limited/1047
Estimation 2S Linear 2S Logit 2S Linear
 Industry,
Effects Industry Industry Year
Dependent Variable ROA dRELATION ROA
dRELATION or -3.08 [***] -5.22 [***]
RELATION (1.02) (1.01)
dSINGLE{0,1,2} 1.27
 (1.08)
dREPLACE{0,1,2} -5.63 [*]
 (3.37)
dSTART{0,1,2} -1.20
 (1.20)
dEND{0,1,2} 0.87
 (1.09)
BANKSHARE
dDnBorCBK
BANKSIZE 0.09
 (0.60)
Goodness of Fit 0.156 0.590 0.218
Log Likelihood -3113.145 -437.607 -3681.658
Panel B
Model/Equation (5.b)
Sample/Number of
Observations
Estimation 2S Logit
 Industry,
Effects Year
Dependent Variable dRELATION
dRELATION or
RELATION
dSINGLE{0,1,2}
dREPLACE{0,1,2}
dSTART{0,1,2}
dEND{0,1,2}
BANKSHARE 2.65
 (1.77)
dDnBorCBK
BANKSIZE
Goodness of Fit 0.650
Log Likelihood -374.48
COPYRIGHT 2001 Financial Management Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2001 Gale, Cengage Learning. All rights reserved.

 
Article Details
Printer friendly Cite/link Email Feedback
Author:Degryse, Hans; Ongena, Steven
Publication:Financial Management
Geographic Code:4EXNO
Date:Mar 22, 2001
Words:14373
Previous Article:Errata.
Next Article:The Initial Listing Decisions of Firms That Go Public.
Topics:


Related Articles
Real-Time Customer Data ...That's Useful for Retention.
Establishing Unique Customer Relations Using Data Warehousing.
Here's How an MCIF Can Pay for Itself.
Making 'Customer profitability' Mark.
Using Profitability Data--Profitably!
Lessons in ACH data. (Database Marketing).
Bank relationships and their effects on firm performance around the Asian financial crisis: evidence from Taiwan.
How to integrate your bank's profit system into the CRM.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters