Bank capitalization and lending behavior after the introduction of the ...

3 downloads 217 Views 243KB Size Report
In this paper, we provide evidence that banks with a low level of capitalization have reduced their commitment with respect to lines of credit after the introduction ...
Rev Quant Finan Acc (2007) 28:147–162 DOI 10.1007/s11156-006-0009-4

Bank capitalization and lending behavior after the introduction of the Basle Accord Ling Chu · Robert Mathieu · Sean Robb · Ping Zhang

Published online: 1 December 2006  C Springer Science + Business Media, LLC 2006

Abstract In this paper, we provide evidence that banks with a low level of capitalization have reduced their commitment with respect to lines of credit after the introduction of the Basle Accord. A bank’s lending behavior reflects its level of commitment towards borrowers, which in turn affects the level of effort it exerts on screening and monitoring the activities of borrowers. We find that the post-Basle Accord market reaction to the announcement of lines of credit issued by banks with a low level of capitalization is significantly lower than the reaction to other types of bank credit announcements. We interpret this result as evidence that some banks have a low level of commitment associated with lines of credit after the Basle Accord. Keywords Bank capitalization . Credit agreements . Basle Accord

1 Introduction We examine the impact of banks’ capitalization on their lending behavior around the Basle Accord. Like most industrial countries, the United States, through the Federal Deposit Insurance Corporation (FDIC), adopted the recommendations of the Basle Committee (the Basle Accord). Since capital adequacy measures are tied to bank credit risk, the adoption of the Basel Accord allows regulators to control the risk behaviour L. Chu · R. Mathieu School of Business and Economics, Wilfrid Laurier University S. Robb () Kenneth G. Dixon School of Accounting, University of Central Florida, P.O. Box 161400, Orlando, FL 32618-1400, USA e-mail: [email protected] P. Zhang Rotman School of Management, University of Toronto Springer

148

L. Chu, R. Mathieu et al.

of banks. When a bank increases its credit risk the total value of its risk-adjusted assets is increased, and consequently, it may have to issue additional equity or undertake other costly actions to comply with minimum capital requirements. While the Basle Accord may prevent some banks from engaging in excessive risk-taking behavior that can be prejudicial to depositors, it may also affect a bank’s lending strategies. For example, by either reducing the length of credit agreements or by introducing a clause that allows them to unconditionally cancel lines of credit at any time, a bank can eliminate the negative (costly) impact of unused commitments on the calculation of its capital ratio. Consequently, it is possible that the introduction of the Basle Accord has reduced banks’ level of commitment at the issuance of lines of credit, which in turn may impact borrowers either by reducing their flexibility in planning future investment opportunities, or by reducing their ability to use a line of credit as an important signaling device to convey private information to investors. Prior to providing a loan to a borrower, a bank investigates the borrower’s payback ability. The existing literature provides evidence that the disclosure of bank credit agreements can provide a signal to the market of a borrower’s ability to raise capital and of a bank’s assessment of the borrower’s value (see, for example, James, 1987; Johnson, 1996; Preece and Mullineaux, 1996; Aintablian and Roberts, 2000). A bank’s assessment of borrower quality depends on both the results of its screening activities and the impact of its monitoring activities through the duration of the loan. As the quality of a bank’s screening and monitoring activities increases, the market reaction to credit agreement announcements should be stronger. If banks changed their screening and monitoring activities following the introduction of the Basle Accord, we should observe a change in the market reaction at the disclosure of bank credit agreements. Prior research (i.e., Andre et al., 2001) examines the markets’ ex ante and ex post reaction to the introduction of the Basle Accord but fails to consider the impact of a bank’s capitalization level on its lending strategy. This omission is crucial since the Basle Accord is concerned with capital adequacy requirements. Our main contribution is to control for banks’ level of capitalization, which enables us to better understand the impact of capital requirements on lending strategies. For instance, at the issuance of a line of credit a bank may not reduce its lending commitment if it believes that its capital ratio is sufficiently high to withstand the negative effect of this new line of credit. However, if the negative impact on the capital ratio can affect the bank’s ability to satisfy minimum capital requirements, the bank may decide to reduce its commitment at the issuance of a line of credit in order to preserve an adequate capital ratio. As a result, the introduction of the Basle Accord may have reduced the informativeness of lines of credit issued by banks with lower capital ratios, but not the informativeness of those issued by banks with higher capital ratios. Our results indicate that the information content conveyed by the disclosure of lines of credit is significantly lower after the introduction of the Basle Accord. However, by controlling for a bank’s level of capitalization we provide evidence that the information content is lowered only when lines of credit are issued by banks with lower levels of capitalization. In other words, our results suggest that not all banks have changed their lending strategies after the introduction of the Basle Accord—only banks with (relatively) low capitalization levels have altered their lending strategy.

Springer

Bank capitalization and lending behavior after the introduction of the Basle Accord

149

The remainder of the paper is organized as follows. We develop our hypotheses in Section 2 and describe the data and methodology in Section 3. Section 4 presents the multivariate analysis. Conclusions are offered in Section 5. 2 Hypotheses development In the early 1990s most industrial countries changed the capital adequacy requirements imposed on financial institutions in order to comply with the recommendations of the Bank of International Settlement.1 The Basle Accord introduced a common definition of capital using a weighting system to calculate the minimum capital requirement and takes into consideration off-balance sheet activities.2 Consequently, under the Basle Accord both used and unused commitments are considered in the denominator of the capital adequacy ratio calculation and their impact on this ratio is dependant on the weight applied to similar instruments. Nonetheless, it is possible for banks to eliminate the impact of unused lines of credit on the capital ratio.3 Like all off-balance sheet instruments, the inclusion of unused lines of credit in the denominator follows a two-step procedure. First, a credit conversion factor (i.e., a weight) is applied to obtain the asset equivalent. Second, the weight applied to similar assets is used to convert these asset equivalents into risk-adjusted off-balance sheet activities. While a credit conversion factor of 50% is typically applied to unused commitments, the regulations also allow a different weight in the following cases: 0%: for unused portions of commitments with an original maturity of one year or less. 0%: for unused portions of commitments (regardless of maturity), which are unconditionally cancelable at any time, provided a separate credit decision is made before each drawing. Consequently, banks can eliminate the negative (costly) impact of unused commitments on the calculation of the capital ratio by either reducing the length of credit agreements to one year or less or by introducing a clause that allows them to unconditionally cancel lines of credit at any time. In either case, the change in the basic conditions of the line of credit reduces the commitment of the bank to its clients. If banks reduce their level of commitment following the Basle Accord (which in turn leads to lower levels of screening and monitoring activities), we should observe: (1) a weaker market reaction to the disclosure of lines of credit after 1991, (2) no difference in the market reaction to the disclosure of term loans and lines of credit prior to 1990, and (3) a weaker market reaction to the disclosure of lines of credit relative to term loans after 1991. These results were documented in Andre et al. (2001) in the 1 In the United States banks had to gradually implement these requirements starting January 1, 1990 and were expected to fully comply by December 31, 1991. 2 Prior to the introduction of the Basle Accord, unused lines of credit had no impact on a bank’s ability to meet minimum capital requirements. 3 When the capital ratio of a bank is close to the minimum required level, regulators monitor the bank more closely. To avoid this scrutiny a bank may, for instance, have to issue additional capital or change the composition of its assets.

Springer

150

L. Chu, R. Mathieu et al.

context of Canadian banks. Given that we use a sample of US firms, we first test these expectations to ensure their robustness in a different economic context.4 Banks are more likely to incur the costs related to maintaining capital adequacy ratios when they are close to the minimum capital requirement (i.e., when they have a low level of capitalization).5 We define a bank’s level of capitalization as the total risk-based capital ratio for the fiscal year prior to the release of the credit agreement.6 We predict that banks with a low level capitalization are more concerned about the impact of unused commitments on their capital ratio than banks with a high level of capitalization after the introduction of Basle Accord and hence, they are more likely to reduce their level of commitment. Therefore, we predict a lower market reaction when credit agreements are provided by banks with a low level of capitalization than when they are provided by banks with a high level of capitalization.7 This leads to the following hypothesis: Hypothesis 1: After 1991, the market reaction to the announcement of lines of credit is positively related to the bank’s level of capitalization. The following two hypotheses compare the market reaction to the disclosure of lines of credit around the introduction of the Basle Accord, after controlling for the level of capitalization post-1991. Andre et al. (2001) provide evidence that the information content conveyed by the disclosure of lines of credit is lower after the introduction of the Basel Accord. They also provide evidence that the disclosure of term loans and lines of credit provide similar information before the Basle Accord, while the informativeness is higher for term loans than for lines of credit after the Basle Accord. We conjecture that banks’ capitalization levels play a role in the lack of information content conveyed by the disclosure of lines of credit post Basle Accord. We predict that not all banks are likely to change the terms of their credit agreements after the introduction of the Basle Accord. More precisely, the likelihood of being in violation of the minimum capital requirement due to the issuance of lines of credit is less significant for banks with a high level of capitalization. For these banks, we do not expect to see a difference in the market reaction at the issuance of lines of credit before and after the Basle Accord. However, since banks with a low level of 4 We use an event study methodology in this paper. We examine the market reaction at the disclosure of bank credit agreements to infer the impact of the Basle Accord on bank lending behavior. Given that the financial press summarizes the credit agreements, we face a data constraint in that we cannot directly observe the introduction of clauses to reduce a bank’s level of commitment. Alternatively, it is possible to study banks’ level of commitment by examining the terms of the loans from the 10-K forms which are available several months after the disclosure of the credit agreements. 5

Liu et al. (1997) define banks that are below (above) the mean total capital ratio as “at risk” (“not at risk”) banks. When we introduce this categorization in the analyses to distinguish between banks with a low or a high level of capitalization we obtain the same conclusions. An alternative test would be to define banks that are under or just at the minimum capital requirement as at risk banks. However, as indicated in Table 1 all of the banks in our sample meet the minimum capital requirements.

6

Bank data is obtained from the Federal Deposit Insurance Corporation website.

7

If a bank wants to change the maturity or terms of their lines of credit, it is possible that borrowers will switch banks. Under this scenario, only the most financially constrained borrowers would be put in this position. We test this possibility by comparing the financial position (e.g., leverage) of firms borrowing from banks with a low level of capitalization to those borrowing from banks with a high level of capitalization and we observe no significant differences between the two groups. Springer

Bank capitalization and lending behavior after the introduction of the Basle Accord

151

capitalization are more likely to be concerned about the impact of lines of credit on their capital ratio, the market reaction for these banks post-1991 should be smaller than the reaction observed prior to 1990. Hypothesis 2: The market reaction to the announcement of lines of credit issued by any bank before 1990 is stronger than the market reaction to the announcement of lines of credit by banks with a low level of capitalization after 1991. Hypothesis 3: The market reaction to the announcement of lines of credit issued by any bank before 1990 is similar to the market reaction to the announcement of lines of credit by banks with a high level of capitalization after 1991. Combining the predictions of Andre et al. (2001) with our Hypotheses 1, 2 and 3, we develop these additional hypotheses: Hypothesis 4: After 1991, the market reaction to the announcement of term loans issued by any bank is stronger than the market reaction to the announcement of lines of credit by banks with a low level of capitalization. Hypothesis 5: After 1991, the market reaction to the announcement of term loans issued by any bank is similar to the market reaction to the announcement of lines of credit by banks with a high level of capitalization.

3 Data, methodology and descriptive statistics Bank credit agreement announcements (866) were obtained from the Wall Street Journal for the period 1980 to 1998.8 Missing stock prices on the CRSP database reduces the sample to 573 announcements. Another 112 firms were excluded due to thin trading problems and 5 firms were excluded due to the absence of prices in the announcement period.9 The final sample consists of 456 observations. Table 1 provides descriptive statistics for the complete period, and where relevant, for the period before and after the introduction of the Basle Accord. As indicated in Panel A, the sample consists of 316 lines of credit, 94 term loans and 46 announcements containing both a line of credit and a term loan. The sample is composed of 218 new credit agreements, 233 revised credit agreements, and 5 combined new and revised credit agreements. Given that the vast majority of the revisions were positive for our sample of firms, we do not control for this additional refinement in our tests.10 Most 8 An exhaustive search was conducted using key words to identify all publicly available announcements of bank credit agreements disclosed in the Wall Street Journal. To ensure that all disclosures were identified in the initial search, we used general expressions such as “loan”, “credit”, “financing”, “credit agreement”, “restructuring”, “commitment”, “refinancing”, “credit line”, and “revolving”. 9 We require at least 100 days of trading in the estimation period in order to calculate expected returns using the market model. 10 Consistent with Lummer and McConnell (1989) and Best and Zhang (1993), the favorable group includes loans for which one of the following occurred: (1) the maturity of the credit agreement is lengthened; (2) the interest rate is reduced; (3) the amount of the loan is increased; or (4) the debt covenants are made less restrictive. The unfavorable group includes loans for which at least one of the above criteria is revised negatively while no terms are revised positively. The mixed group contains loans having at least one term revised favorably and one term revised unfavorably. A total of 195 revisions were favorable, 16 were unfavorable and 22 were mixed.

Springer

152

L. Chu, R. Mathieu et al.

Table 1

Descriptive statistics

Full sample

Number of observations 1980–1989 1990–1991 1992–1998 456 176 56 224

(A) Sample composition Type of credit agreements Lines of credit 316 107 Term loans 94 45 Lines of credit and term loans 46 24 New versus revised credit agreements New credit agreements 218 88 Revised credit agreements 233 84 New and revised credit agreements 5 4 Number of lenders Credit agreements provided by single banks 108 38 Credit agreements provided by multiple banks 311 129 Information not provided 37 9 (B) Loan amount, firm size and leverage ratio Number of observations Mean

Median

Loan amount (in millions of dollars) 443 519 110 309 506 122 88 562 101 46 533 77 216 619 110 222 432 115 Firm size (in million of dollars) Full sample 436 2,951 437 (C) Banks’ characteristics (after December 31, 1991)b Adequacy ratios and non-performing loans Total capital ratio (in%) 125 11.81 11.53 Tier 1 capital ratio (in%) 125 8.67 8.10 Ratio of non-performing loans 125 2.76 1.69 over net loans (in%) Net loans and deposits (in millions) Net loans (used) 125 72,339 61,742 Total deposits 125 80,837 62,274 Full samplea Lines of credit Term loans Lines of credit and term loans New agreements Revised agreements

a This

37 16 3

172 33 19

20 36 0

110 113 1

9 43 4

61 139 24

Minimum

Maximum

0.3 2.5 0.3 0.5 0.3 2.5

20,600 20,600 11,000 7,600 20,600 7,600

2

191,013

8.82 6.68 0.22

21.86 21.19 14.62

59 80

195,565 198,317

represents the total number of observations for which information is available.

b There

are 280 announcements made after January 1, 1990. A total of 83 observations are eliminated because the lending banks cannot be identified. We also eliminate a total 45 credit agreements that are provided by foreign banks and an additional 27 announcements made in 1990 and 1991. The remaining 125 announcements consist of 99 lines of credit and 26 term loans.

credit agreements (311) are provided by a syndicate of banks, a single bank is involved in 108 agreements and information is not available about the specific lenders for 37 credit agreements. Descriptive statistics regarding the size of the credit agreements are provided in Panel B. The average amount is $519 million and the mean is slightly higher for term loans than for lines of credit. Springer

Bank capitalization and lending behavior after the introduction of the Basle Accord

153

Panel C provides summary characteristics of the lending banks.11 A total of 280 bank credit agreements are disclosed after January 1, 1990. From this total, 83 press releases do not mention the name of the lenders. From the 197 announcements where the name of the lenders are mentioned, 45 credit agreements are provided by banks that are not members of the Federal Deposit Insurance Corporation (FDIC).12 As a result, we are only able to obtain bank characteristics for a total of 152 announcements. We present statistics for agreements disclosed after 1991 (a total of 125 observations). As indicated in Panel C, all the banks included in our sample after January 1, 1992 have a capital ratio that exceeds the minimum requirement of 8% (mean of 11.81% and median of 11.53%), with a minimum (maximum) ratio of 8.82% (21.86%). The average ratio of non-performing loans over net loans is 2.76% (median of 1.69%), with a minimum (maximum) ratio of 0.22% (14.62%). The second part of Panel C provides information about net loans provided by the 125 banks as well as the total value of deposits at these financial institutions. On average, our sample banks provided $72,339 million in net loans and the mean ratio of loans over deposits is 89.5%. To analyze the information content of bank credit agreements we employ an event study methodology to examine the relationship between changes in a firm’s market value at the announcement of bank credit agreements. As in Brown and Warner (1980), James (1987), and Lummer and McConnell (1989), excess returns are calculated using the market model with the two-day event window defined as the day of the announcement in the Wall Street Journal (t = 0) and the previous day (t = −1). The parameters of the market model are estimated over the period t = −240 to t = −11 prior to the announcement using daily returns. Tests of statistical significance (z statistics) of the average abnormal returns are based on standardized prediction errors using the parameters of the market model (see James, 1987). Only firms with a minimum of 100 daily observations are kept in the analysis. Table 2 provides the market reaction at the disclosure of bank credit agreements. As indicated in Panel A, the market reacts positively to the disclosure of bank credit agreements (average announcement excess return of 1.73 percent and z statistic of 6.03). Therefore, consistent with prior studies, the market seems to infer the bank’s assessment of a firm’s quality at the disclosure of bank credit agreements (see, for example, James, 1987; Johnson, 1996; and Aintablian and Roberts, 2000).13 The null hypothesis that the market reactions are the same for term loans versus lines of credit, for new versus revised credit agreements, for credit agreements provided by multiple versus single banks and for credit agreements provided to small versus large firms cannot be rejected at conventional levels. Panel B of Table 2 examines the market reaction at the disclosure of lines of credit and term loans for the period before and after the introduction of the Basle Accord.14 Given that banks had to gradually implement the recommendations of the Basle Accord between January 1, 1990 and December 31, 1991, we eliminate the years 11 Consistent with Billett et al. (1995), when a syndicate of banks is involved we use the descriptive statistics of the lead bank. 12

The non-members of the FDIC are foreign (i.e., non-US) banks.

13

See Boot (2000) and Ongena and Smith (2000) for a review of the literature.

14

We combine announcements of term loans, with announcements involving both a line of credit and a term loan, into one category. Springer

154 Table 2

L. Chu, R. Mathieu et al. Average announcement excess returns Number of observations

Full sample Lines of credit

Announcement period excess returns (%)

Panel A: Characteristics of bank credit agreements and firm size 456 1.73 316 1.09

z-statistica 6.03∗∗∗ 3.54∗∗∗

Term loans Term loans & lines of credit New credit agreements

94 46 218

3.33 2.83 1.36

5.35∗∗∗ 2.05∗∗ 4.26∗∗∗

Revised credit agreements Credit agreements provided by a single bank Credit agreements provided by multiple banksb

233 108

2.11 2.30

4.29∗∗∗ 2.79∗∗∗

348

1.55

5.35∗∗∗

Small firms 218 1.49 3.40∗∗∗ Large firms 218 1.89 4.78∗∗∗ Panel B: The informativeness of lines of credit and term loans before 1990 and after 1991

Lines of credit Term loans

Obs.

Before 1990 Excess returns z-statistic

107 69

2.50 1.12

t-statistic (lines of credit versus term loans) a ∗∗∗ Significant

0.83

5.77∗∗∗ 2.25∗∗

Obs. 172 52

After 1991 Excess returns z-statistic 0.08 2.70

− 0.80 2.49∗∗

t-statistic Before versus after 2.39∗∗ 0.73

−1.79∗

at 0.01 level,∗∗ significant at 0.05 level, ∗ significant at 0.10 level.

b

This category includes credit agreements provided by multiple banks and those for which the number of lending banks cannot be determined from the announcement.

1990 and 1991 from our tests. The results are consistent with our hypotheses. More precisely, we observe: (1) a stronger market reaction to the disclosure of lines of credit before 1990 than after 1991 (t-statistic of 2.39); (2) no difference between the market reactions at the disclosure of lines of credit and term loans prior to 1990 (t-statistic of 0.83); and (3) a stronger market reaction to the disclosure of term loans relative to lines of credit after 1991 (t-statistic of −1.79). For completeness, we also examine the market reaction to the disclosure of term loans around the implementation of the Basle Accord and the difference between the two market reactions is not significant at conventional levels (t-statistic of 0.73).

4 Multivariate analysis 4.1 The informativeness of lines of credit and term loans before 1990 and after 1991 In this section, we test the relationship between announcement period excess returns and the variables used in the univariate analysis to examine the market reaction to the Springer

Bank capitalization and lending behavior after the introduction of the Basle Accord

155

disclosure of bank credit agreements. The use of a regression enables us to estimate the effects of the characteristics of bank credit agreements on excess returns in a controlled environment. As in Lummer and McConnell (1989), Johnson (1996) and Andre et al. (2001), among others, we control for heteroscedasticity in cross-sectional stock returns by using a weighted least squares regression employing the inverse of the relevant standard prediction errors as the weight. The first analysis examines the market reaction to the disclosure of lines of credit before 1990 and after 1991. We control for characteristics of bank credit agreements that are typically available in press releases: new versus revised credit agreements and concentration of borrowing (single versus multiple lenders). The impact of the credit agreement on the market reaction at the time of the announcement may vary with the borrowing firm’s existing debt characteristics, therefore, we control for leverage.15 In addition, we include the relative loan amount in the regression. In Table 3, we estimate the following regression (the subscript i indicates the borrowing firm): PEi = α + β1 DATE + β2 NEW REVISEi + β3 BANK NUM i +β4 LEVERAGE + β5 REL LOAN + εi

(1)

where: PE is the two-day excess return; DATE is a dummy variable that takes the value of 1 when a firm receives a credit agreement issued after 1991 and 0 when issued before 1990; NEW REW is a dummy variable that takes the value of 1 for a revised credit agreement and 0 for a new credit agreement; BANK NUM is a dummy variable that takes the value of 1 when the credit agreement is provided by a single bank and takes the value of 0 when it is provided by multiple banks (or when the number of banks cannot be determined); LEVERAGE is the firm’s leverage ratio (long term debt divided by total assets from the previous annual report); REL LOAN is the amount provided in the credit agreement divided by total assets; and εi is the noise term. Predicted signs are provided in Column A of Table 3. The variable DATE captures the impact of the Basle Accord and we predict a negative sign (see Andre et al., 2001). The sign of NEW REV is expected to be positive (see, for example, Lummer and McConnell, 1989 and Best and Zhang, 1993). The sign of BANK NUM is indeterminate. Petersen and Rajan (1994) provide evidence that concentration of borrowing is viewed as good news. Preece and Mullineaux (1996) also provide evidence that the market reaction is a declining function of the number of lending banks. However, Rajan (1992) claims that the information acquired by a bank can create an “information monopoly” or hold-up problem in that it is costly for the borrower to switch lenders. Houston and James (1996) support Rajan’s claim by providing evidence that firms borrowing from multiple banks undertake more investment opportunities than firms borrowing from a single bank because, in the latter case, the firm does not have incentives to invest in new projects given that the bank uses its information monopoly to capture most of the profits. If this is true, the sign of BANK NUM could be negative. The sign of LEVERAGE is expected to be positive, since Johnson (1996) provides evidence that more leveraged firms benefit more from the announcements of credit agreements. Finally, the sign of REL LOAN is indeterminate. On one hand, 15 We also estimate Eq. (1) after controlling for firm size. Consistent with the univariate analysis, our results are not affected by the inclusion of a size variable.

Springer

156

L. Chu, R. Mathieu et al.

Table 3

The market reaction to the disclosure of lines of credit Column

Lines of credit

Independent variables

(A) Sign

(B) Coefficient

(C) t statistica

Intercept DATE NEW REV BANK NUM LEVERAGE REL LOAN

? − + ? + ?

0.0193 − 0.0323 0.0257 − 0.0148 0.0203 − 0.0010

1.35 − 2.67∗∗∗ 2.20∗∗ − 1.16 1.34 − 1.61

Observations Adjusted R2 (%)

264 4.63

Variables: DATE: a dummy variable that takes the value of 1 when a firm receives a credit agreement issued after 1991 and 0 when issued before 1990; NEW REV: a dummy variable that takes the value of 1 for a revised credit agreement and 0 for a new credit agreement; BANK NUM: a dummy variable that takes the value of 1 when the credit agreement is provided by a single bank and takes the value of 0 when it is provided by multiple banks (or when the number of banks cannot be determined); LEVERAGE: the firm’s leverage ratio (long term debt divided by total assets); and REL LOAN: the amount provided in the credit agreement divided by total assets. a ∗∗∗

significant at 0.01 level, ∗∗ significant at 0.05 level, ∗ significant at 0.10 level.

the loan amount signals the bank’s willingness to finance the firm’s activities. On the other hand, it reflects an increase in equity risk. Results from estimating Eq. (1) are presented in Table 3. The sign on the variable DATE is negative, as expected, and significant (at the 1% level), implying that the market reaction to the disclosure of lines of credit is weaker after 1991 than before 1990. The control variable NEW REV is also significant and has the expected sign. To further examine the impact of the Basle Accord, we modify Eq. (1) by replacing the variable DATE with the variable LOC TL: PEi = α + β1 LOC TL + β2 NEW REVISEi + β3 BANK NUM i +β4 LEVERAGE + β5 REL LOAN + εi

(2)

where: LOC TL is a dummy variable with a value of 1 when a firm receives a term loan and 0 otherwise. All other variables are as previously defined. The variable LOC TL captures the difference in the information content of lines of credit and term loans. In Table 4, Columns (B) and (C) present the results for the period prior to 1990. We hypothesize that there is no difference between the market reaction to the disclosure of term loans and lines of credit prior to the introduction of the Basle Accord. Consistent with this belief, the coefficient on LOC TL is insignificant (t-statistic of −0.94). Columns (D) and (E) of Table 4 present the results for the period after 1991. We predict a stronger market reaction to the disclosure of term loans, relative to lines of Springer

Bank capitalization and lending behavior after the introduction of the Basle Accord Table 4

157

The market reaction to the disclosure of term loans and lines of credit

Column (A) Independent variables Sign

(B) Results

Intercept LOC TL NEW REV BANK NUM LEVERAGE REL LOAN

0.0224 − 0.0234 0.0421 − 0.0310 0.0037 − 0.0011

? ?, + + ? + ?

Observations Adjusted R2 (%)

Before 1990 (C) t statistica 1.00 − 0.94 1.76 − 1.17 0.84 − 1.14

(D) Results − 0.0021 0.0469 − 0.0006 0.0229 0.0279 − 0.0278

154 1.38

After 1991 (E) t statistic − 0.12 2.89∗∗∗ − 0.04 1.46 1.11 − 1.70 ∗ 212 5.79

Variables: LOC TL: a dummy variable that takes the value of 1 when a firm receives a term loan and takes the value of 0 otherwise; NEW REV: a dummy variable that takes the value of 1 for a revised credit agreement and 0 for a new credit agreement; BANK NUM: a dummy variable that takes the value of 1 when the credit agreement is provided by a single bank and takes the value of 0 when it is provided by multiple banks (or when the number of banks cannot be determined); LEVERAGE: the firm’s leverage ratio (long term debt divided by total assets); and REL LOAN: the amount provided in the credit agreement divided by total assets. a ∗∗∗ significant

at 0.01 level, ∗∗ significant at 0.05 level, ∗ significant at 0.10 level.

credit, after the introduction of the Basle Accord. The positive sign and the significant coefficient, at the 1% level, on the variable LOC TL are consistent with our prediction (t-statistic of 2.89). 4.2 The impact of banks’ level of capitalization on the informativeness of lines of credit In this section we examine the impact of banks’ level of capitalization on the market reaction to the disclosure of lines of credit after the introduction of the Basle Accord. Consistent with the previous analysis, we control for heteroscedasticity in crosssectional stock returns by using a weighted least squares regression with the inverse of the relevant standard prediction errors as the weight. In the following models (Eqs. (3) and (4)), we add variables to measure banks’ capital ratios and non-performing loans.16 To test Hypothesis 1, we estimate the following model in Table 5: PEi = α + β1 TOTAL RATIO + β2 NEW REVISEi + β3 BANK NUM i +β4 LEVERAGE + β5 NPL + β6 REL LOAN + εi

(3)

16 The amount of non-performing loans is not included in Eqs. (1) and (2) since the information is not available on the FDIC website for the entire sample period covered by these tests.

Springer

158

L. Chu, R. Mathieu et al.

Table 5 The market reaction to the disclosure of lines of credit after 1991—controlling for bank capitalization (a test of Hypothesis 1) Column

Columns

(A) Independent variables Sign

(B) Coefficient

(C) t statistica

Intercept TOTAL RATIO NEW REV BANK NUM LEVERAGE NPL REL LOAN

− 0.1369 0.0109 0.0186 − 0.0367 − 0.0024 0.8726 − 0.0163

− 2.34 2.27∗∗ 1.07 − 1.97 ∗ − 0.08 3.06∗∗∗ − 0.93

? + + ? + + ?

Observations Adjusted R2 (%)

92 18.28

Variables: TOTAL RATIO: the bank’s total capital adequacy ratio; NEW REV: a dummy variable that takes the value of 1 for a revised credit agreement and 0 for a new credit agreement; BANK NUM: a dummy variable that takes the value of 1 when the credit agreement is provided by a single bank and takes the value of 0 when it is provided by multiple banks (or when the number of banks cannot be determined); LEVERAGE: the firm’s leverage ratio (long term debt divided by total assets); NPL: the ratio of non-performing loans to net loans for the lead bank; and REL LOAN: the amount provided in the credit agreement divided by total assets. a ∗∗∗

significant at 0.01 level, ∗∗ significant at 0.05 level, ∗ significant at 0.10 level.

where: TOTAL RATIO is the lead bank’s total capital ratio; and NPL is the ratio of non-performing loans over net loans for the lead bank. All other variables are as previously defined. We use as a sample the announcements of lines of credit made after 1991. The variable TOTAL RATIO captures the market reaction to the disclosure of lines of credit issued by banks with specific levels of capitalization.17 According to Hypothesis 1, we predict a positive sign on this variable.18 We also include the ratio of non-performing loans over total (net) loans, NPL. We expect banks to become stricter in their evaluation process when they experience high levels of loan default since it negatively affects earnings. As a result, we expect that loan quality will be increasing in the level of non-performing loans as 17 An examination of the annual distribution of banks’ capital ratios over our sample period does not reveal any systematic patterns in the average annual values. 18

Billett et al. (1995) provide evidence that a bank’s credit quality conveys information to capital markets. They observe a positive association between the quality of a bank’s credit rating and the magnitude of the market reaction to the announcement of a credit agreement. To examine this possibility, we collect the available Moody’s credit ratings for our sample banks’ senior non-secured debt. The correlation between the level of bank capitalization and credit ratings is low (−0.27) implying that credit ratings may provide additional explanatory power in our tests. To examine this possibility, we code the Moody’s ratings in a manner consistent with Billett et al. (1995) and include a credit rating variable in Eq. (3). The results (not reported) indicate that the credit rating variable is insignificant and its inclusion does not affect our main conclusions from Table 5. Springer

Bank capitalization and lending behavior after the introduction of the Basle Accord

159

banks attempt to reduce their overall loan default rate. Consequently, the sign on this variable should be positive.19 Columns B and C of Table 5 present the results from estimating Eq. (3). Consistent with Hypothesis 1, the coefficient on the variable TOTAL RATIO is positive, as expected, and significant at a 5% level (t-statistic of 2.27). The variable NPL is significant at a 1% level and has the expected sign. To test Hypotheses 2, 3, 4 and 5, we re-estimate Eq. (3) after substituting a unique dummy variable in place of LOC TL in order to test each prediction. We estimate the following model(s) in Tables 6 and 7:20 PEi = α + β1 H ( j) + β2 NEW REVISEi + β3 BANK NUM i +β4 LEVERAGE + β5 REL LOAN + εi

(4)

where: H(j) is one of: Hypotheses (j = 2, 3, 4, or 5), and H(2) is a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a low level of capitalization and 0 when a line of credit is issued before 1990 (by any bank); H(3) is a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a high level of capitalization and 0 when a line of credit is issued before 1990 (by any bank); H(4) is a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a low level of capitalization and 0 when a term loan is issued after 1991 (by any bank); and H(5) is a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a high level of capitalization and 0 when a term loan is issued after 1991 (by any bank). All other variables are as previously defined. Recall that from Hypotheses 2 and 3, we believe that the market reaction to the announcement of lines of credit issued by any bank before 1990 should be stronger than (similar to) the market reaction when lines of credit are issued by banks with a low (high) level of capitalization after 1991. Consequently, we predict a negative sign on H(2), while the coefficient on H(3) should be insignificant. Consistent with Liu et al. (1997), we distinguish banks with a low and a high level of capitalization as those below and above the mean capitalization level. Table 6 reports results from estimating Eq. (4) that supports these predictions. From Hypotheses 4 and 5, we believe that after 1991, the market reaction to the announcement of term loans issued by any bank should be stronger than (similar to) the market reaction when lines of credit are issued by banks with a low level (high

19 Alternatively, if a bank has a large amount of non-performing loans, it may indicate that the bank’s screening and monitoring activities are not effective. As a result, the market may not positively react to the announcement of a credit agreement provided by such a bank. In this case, we would observe a negative sign on NPL. See Liu and Ryan (1995) for a discussion on the signaling role of NPL. 20 The variable NPL is not included in Tables 6 and 7. Table 6 compares the market reaction to the disclosure of lines of credit before and after the introduction of the Basle Accord. Given that the value of non-performing loans is not available for the entire sample period, we exclude this variable. Table 7 compares term loans (issued by any bank) to lines of credit (issued by a bank with a low level of capitalization and banks with a high level of capitalization). To maximize the number of observations, we do not limit our analysis to the term loans for which the name of the bank is mentioned.

Springer

160

L. Chu, R. Mathieu et al.

Table 6 The market reaction to the disclosure of lines of credit before 1990 and after 1991—controlling for bank capitalization (a test of Hypotheses 2 and 3) Column

Independent variables

(A) Sign

Intercept H(2) H(3) NEW REV BANK NUM LEVERAGE REL LOAN

? − ? + ? + ?

Observations Adjusted R2 (%)

Lines of credit (Low capitalization) (B) (C) Results t statistica 0.0252 − 0.0363

1.49 −2.19∗∗

0.0398 −0.0360 0.0041 − 0.0012

2.43∗∗ −2.11∗∗ 0.23 −1.71 ∗

Lines of credit (High capitalization) (D) (E) Results t statistic 0.0200

1.11

−0.0043 0.0266 −0.0089 0.0115 −0.0010

− 0.21 1.47 − 0.46 0.57 − 1.46

158 8.43

138 0.00

Variables: H(2): a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a low level of capitalization and 0 when a line of credit is issued before 1990 (by any bank); H(3): a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a high level of capitalization and 0 when a line of credit is issued before 1990 (by any bank); LOC TL: a dummy variable that takes the value of 1 when a firm receives a term loan and takes the value of 0 otherwise; NEW REV: a dummy variable that takes the value of 1 for a revised credit agreement and 0 for a new credit agreement; BANK NUM: a dummy variable that takes the value of 1 when a credit agreement is provided by a single bank and takes the value of 0 when it is provided by multiple banks (or when the number of banks cannot be determined); LEVERAGE: the firm’s leverage ratio (long term debt divided by total assets); and REL LOAN: the amount provided in the credit agreement divided by total assets. a ∗∗∗ significant

at 0.01 level, ∗∗ significant at 0.05 level, ∗ significant at 0.10 level.

level) of capitalization. We predict a negative sign on H(4) and expect the coefficient on H(5) to be insignificant. Table 7 reports results from estimating Eq. (4) that supports these predictions.

5 Conclusion In this paper, we provide evidence to support the belief that banks with a low level of capitalization engage in reduced levels of screening and monitoring activities when they issue lines of credit after the introduction of the Basel Accord. Our conclusion is based on an examination of the market reaction at the announcement of bank credit agreements. A bank’s assessment of borrower quality depends on both the results of its initial screening activities and on its monitoring activities over the duration of the loan. As the quality of a bank’s screening and monitoring activities increases, the market reaction to credit agreement announcements should be stronger. If banks changed their screening and monitoring activities following the introduction of the Springer

Bank capitalization and lending behavior after the introduction of the Basle Accord

161

Table 7 The market reaction to the disclosure of term loans vs. lines of credit after 1991—controlling for bank capitalization (a test of Hypotheses 4 and 5)

Independent variables

Column (A) Sign

Intercept H(4) H(5) NEW REV BANK NUM LEVERAGE REL LOAN

? − ? + ? + ?

Observations Adjusted R2 (%)

Term loans versus lines of credit (Low capitalization) (B) (C) Results t statistica 0.0723 −0.0530

2.25∗∗ −2.20∗∗

−0.0065 0.0272 −0.0170 −0.0450

− 0.27 1.10 − 0.41 − 1.42

Term loans versus lines of credit (High capitalization) (D) (E) Results t statistic 0.0465 − 0.0076 − 0.0284 0.0763 − 0.0165 − 0.0237

106 5.26

1.32 − 0.27 −1.06 2.72∗∗∗ −0.32 −0.89 86 8.74

Variables: H(4): a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a low level of capitalization and 0 when a term loan is issued after 1991 (by any bank); H(5): a dummy variable that takes the value of 1 when a line of credit is issued after 1991 by a bank with a high level of capitalization and 0 when a term loan is issued after 1991 (by any bank); LOC TL: a dummy variable that takes the value of 1 when a firm receives a term loan and takes the value of 0 otherwise; NEW REV: a dummy variable that takes the value of 1 for a revised credit agreement and 0 for a new credit agreement; BANK NUM: a dummy variable that takes the value of 1 when the credit agreement is provided by a single bank and takes the value of 0 when it is provided by multiple banks (or when the number of banks cannot be determined); LEVERAGE: the firm’s leverage ratio (long term debt divided by total assets); and REL LOAN: the amount provided in the credit agreement divided by total assets. a ∗∗∗ significant

at 0.01 level, ∗∗ significant at 0.05 level, ∗ significant at 0.10 level.

Basle Accord, we would expect to observe a change in the market reaction to the announcement of bank credit agreements. Our results indicate that the information content conveyed by the disclosure of lines of credit is significantly lower after the introduction of the Basle Accord. However, by controlling for a bank’s level of capitalization we provide evidence that the information content is only lowered when lines of credit are issued by banks with lower levels of capitalization. In other words, our results suggest that not all banks have changed their lending strategies after the introduction of the Basle Accord—only those banks with (relatively) low capitalization levels have altered their strategy.

Acknowledgments The authors gratefully acknowledge financial support from a Social Sciences and Humanities Research Council (SSHRC) Standard Research Grant and from a grant partly funded by Wilfrid Laurier University Operating funds and partly by a SSHRC Institutional Grant awarded to Wilfrid Laurier University. We thank Jean Bedard, Matthew Billet, Daniel Coulombe, Ole-Kristian Hope, Kiridaran Kanagaretnam, Suzanne Paquette and workshop participants at Laval University for their comments and suggestions. Springer

162

L. Chu, R. Mathieu et al.

References Aintablian S, Roberts GS (2000) A note on market response to corporate loan announcements in Canada. J Banking Fin 24:381–393 Andre P, Mathieu R, Zhang P (2001) A note on capital adequacy and the information content of term loans and lines of credit. J Banking Fin 25:431–444 Best R, Zhang H (1993) Alternative information sources and the information content of bank loans. J Fin 48:1507–1522 Billett MT, Flannery MJ, Garfinkel JA (1995) The effect of lender identify on a borrowing firm’s equity return. J Fin 50:699–718 Boot AWA (2000) Relationship banking: What do we know. J Fin Intermed 9:7–25 Brown SJ, Warner JB (1980) Measuring security price performance. J Fin Econ 8:205–258 Houston J, James C (1996) Bank information monopolies and the mix of private and public debt claims. J Fin LI:1863–1889 James C (1987) Some evidence of the uniqueness of bank loans. J Fin Econ 19:217–235 Johnson SA (1996) The effect of bank reputation on the value of bank loans agreements. J Acc, Aud Fin 12(1):83–100 Liu CC, Ryan SG (1995) The effect of bank loan portfolio composition on the market reaction to an anticipation of loan loss provisions. J Acc Res 33(1):77–94 Liu CC, Ryan SG, Wahlen JM (1997) Differential valuation implications of loan loss provisions across banks and fiscal quarters. Acc Rev 72(1):133–146 Lummer SL, McConnell JJ (1989) Further evidence on the bank lending process and the capital-market response to bank loan agreements. J Fin Econ 25:99–122 Ongena S, Smith DC (2000) Bank relationships: A review. In: Zenios SA, Harker P (eds) Performance of financial institutions. Cambridge Press Petersen MA, Rajan RG (1994) The benefits of lending relationships: Evidence from small business data. J Fin 49:3–37 Preece D, Mullineaux DJ (1996) Monitoring, loan renegotiability, and firm value: The role of lending syndicates. J Banking Fin 20:577–593 Rajan RG (1992) Insiders and outsiders: The choice between informed and arm’s-length debt. J Fin 47:1367–1400

Springer