Journal of ELSEVIER
Journal of Banking & Finance 18 (1994) 1155-1176
The overall gains from large bank mergers Joel F. Houston *, Michael D. Ryngaert College of Business Administration, Department of Finance, University of Florida, Gainesville, FL 3261 l-201 7, USA
Received August 1992; final version received June 1993
Abstract We demonstrate that the overall gains (the weighted average of gains to the bidder and target firms) from a recent sample of bank mergers are slightly positive, but statistically indistinguishable from zero. This lends support to recent studies which fail to find any significant cost savings resulting from bank mergers. We also demonstrate the characteristics of mergers that the market perceives as most valuable. These attributes include high prior levels of profitability for the bidder, considerable operations overlap between the target and the bidder, and a method of financing that reveals positive information about the bidder or the synergies likely to be created by the merger. Keywords:
Bank mergers;
JEL classification:
Mergers and acquisitions
G21; G34
1. Introduction The banking industry has undergone considerable consolidation in recent years. This consolidation has included a large number of mergers between large banking institutions.
’ Many
argue
that these
mergers
reflect
market
pressures
to reduce
* Corresponding author. We would like to thank John Race for helpful research assistance and acknowledge the many helpful comments and suggestions that were made by Chris James and an anonymous referee. 1Boyd and Graham (1991) document the industry’s consolidation during the last half of the 1980s. They indicate that the number of banks dropped 15% from 1984 to 1990. They also present evidence that average bank size has increased over this period. 0378-4266/94/$07.00 0 1994 Elsevier Science B.V. AR rights reserved SSDI 0378-4266(94)00034-4
1156
J.F. Houston, M.D. Ryngaert / Journal of Banking & Finance 18 (1994) 1155-l 176
costs and increase efficiency. Others suggest that these mergers are not likely to improve the profitability of merging firms, but instead are pursued by empire building managers at the expense of shareholders. 2 Skepticism about the benefits of large-scale bank mergers is fueled by two distinct lines of research. First, Boyd and Graham (1991) note that most studies fail to find economies of scale in the banking industry for firms with deposits above $100 million. Also, large banks are not the most profitable banks. This suggests that making large banks larger will not create more efficient and profitable banks. Second, recent studies of actual mergers between banking institutions with assets in excess of $100 million find little evidence of a decrease in operating expenses arising from the mergers [e.g., Srinivasan and Wall (1992), Berger and Humphrey (1992) and Srinivasan (1992)]. A notable exception to the above studies is Comett and Tehranian (19921, who examine the post-acquisition performance of thirty large merged banks that combined during the period 1982-1987. They report that the return on equity for merged banks outperforms that of the banking industry in the three years after merger. Even in this study, however, superior performance does not extend to the more traditional measure of performance, return on bank assets. 3 While studies that rely on accounting data to measure the ex post performance of bank mergers are enlightening, accounting numbers immediately following a merger may not fully reflect changes in shareholder wealth. For instance, in the short mn banks may not appear to cut costs (and raise returns) because there are large up-front costs associated with a merger. Long-term cost savings may not be realized for several years. To provide additional insights into the value of recent bank mergers, this paper examines the stock market’s perception of bank mergers in the period 19851991. An efficient stock market provides an unbiased assessment of the gains that may result from combining the two banks. Since stock prices reflect the market’s assessment of all future cash flows, they provide an immediate estimate of the likely gains or losses from each merger. Using stock return data we address two issues. First, does the stock market positively revalue large banking firms that announce an intent to merge? Second, what factors are associated with positive revaluations of merging banks?
’ Bank managers may be willing to undertake value reducing acquisitions if they believe that no mechanism is in place to discipline them for such actions. Mitchell and Lehn (1990) find that industrial firms that make bad acquisitions are more likely to become the target of a hostile takeover than firms that make good acquisitions. Gorton and Rosen (1992), however, provide arguments for why bank managers may be more entrenched than nonbank mergers. 3 Spindt and Tahran (1992) also find that a 1986 sample of merging banks become more profitable after their merger. Their sample, however, is largely composed of smaller banks operating in the range where economies of scale are important. Thus, their results may not apply to mergers between large banks.
J.F. Houston, M.D. Ryngaert/Journal
ofBanking & Finance 18 (1994) 1155-1176
1157
Previous research on the stock market reaction to bank merger announcements concentrates on the stock market revaluation of bidders and/or targets but does not consider the revaluation of the combined bidder and target. While all existing bank merger studies find positive target bank stock returns at the announcement of a merger, many recent studies find negative bidder returns. Thus, it is unclear whether the combined revaluation of the two firms is positive or negative. By focusing on the total return (the return to a value weighted portfolio of the bidder and target) at the time of the merger announcement, we address the question of whether the market believes that bank acquisitions are value-enhancing. We find that the average total return to a completed bank merger is slightly greater than zero at the merger’s announcement, though not significantly different from zero. Positive returns to targets are essentially offset by negative returns to bidders. 4 Our results are consistent with accounting-based studies that illustrate limited efficiency gains from large bank mergers. The negative returns to bidders, are also consistent with bidding firm managers, on average, not being driven by value maximization. It is interesting to note, however, that total merger returns in the latter portion of our sample are positive and significant. Next, we examine the factors influencing the total returns to bank mergers. We first focus on whether the returns to bank mergers are related to the recent past performance of both the acquiring and target firm banks. Berger and Humphrey (1992) argue that mergers where efficient banks acquire inefficient banks could result in cost savings through elimination of “X-inefficiencies”. This suggests that acquiring banks should, on average, be more profitable than target banks and that the revaluation of the combined bidder and target stock should be positively related to the profit gap between the bidder and the target. While bidding banks are more profitable than target banks, differences in profitability between the bidder and target do not explain total abnormal returns across bank mergers. We do, however, find that total abnormal returns are significantly higher when the acquiring bank has been more profitable. This result suggests that banks with good track records are perceived as more likely to engage in acquisitions that are value-increasing, even if the target bank already has a strong performance record. Merck et al. (1990) find similar results for a sample of industrial mergers. We also examine whether “in-market” mergers are more profitable than those where the bidder and target operate in relatively distinct markets. In-market mergers may be more profitable because they allow for the closing of redundant branch and head office facilities. Consistent with this view, we find that total merger returns are positively related to the degree of overlap between bidder and target firms’ operations.
4 Our results contrasts with Bradley et al. (1988) and Stulz et al. (1990) who find significantly positive total gains from nonbank mergers.
large,
11.58
J.F. Houston, M.D. R~n~aert/Jour~al
o~Banki~~ & Finance 18 (1994) 1155-1176
In assessing the market’s reaction to a merger announcement, we also controlled for the mode of acquisition and the relative size of the participants. We find that the market responds positively when a mode of acquisition is selected which signals that the bidding firm has positive private information. Holding all else constant, we also find a significant positive relationship between total merger returns and the relative size of the merger participants. Interestingly, we also find that some variables that explain target or bidder returns, in isolation, do not explain the total returns to mergers and vice versa. This indicates that attempting to make cross-sectional inferences about merger gains from regression analysis of bidder or target returns alone can be misleading. The paper proceeds as follows. Section 2 provides a brief review of the stock market based studies of bank merger gains. Section 3 describes our sample of bank mergers. Sections 4 and 5 present our empirical results. Section 6 concludes.
2. Review of the previous
literature
A number of studies document the returns to bidding and target firms at the announcement of bank mergers. Every study that we are aware of documents positive returns to targets in bank mergers, e.g. Hawawini and Swary (1990) and Comett and De (1991). The evidence for bidders, however, is mixed. Desai and Stover (19851, James and Weir (1987) and Comett and De (1991) document positive abnormal returns to bidding firms in banking acquisitions. However, Neely (19871, Hawawini and Swary (19901, Houston and Ryngaert (1992), and Comett and Tehranian (1992) report negative returns to bidders. Closer inspection of these studies suggests that samples that emphasize larger acquisitions are more apt to find negative bidder returns. 5 Given these results, it is unclear if bank mergers lead to positive market revaluation of the combined bidder and target. To our knowledge, no study has fully examined the overall gains from bank mergers. For a sample of mergers between 1972 and 1987, Hawawini and Swary (1990) find that a weighted average of the average gains (losses) to bidding and target firms is positive. Hawawini and Swary’s measure of total revaluation is: (v,/V,P=,
+ AR,
(1)
where (x/V,) is the pre-merger average ratio of target book to bidder book equity, AR, is the average abnormal return to the targets and AR,, is the average abnormal return to the bidders. They conclude that their results provide tentative evidence that bank mergers create wealth. However, they acknowledge that a more
5 Hawawini and Swary (1990) and Houston and Ryngaert (1992) both find that bidder returns are inversely related to the ratio of the target’s equity value to the bidder’s equity value.
J.F. Houston, M.D. Ryngaert/Journal
ofBanking & Finance 18 (1994) 1155-1176
1159
complete analysis would calculate the gain from each merger and then aggregate across all mergers. Their results also do not include any post-1987 mergers. Our second goal is to explain the factors influencing total merger returns. Again, virtually all studies that examine cross-sectional differences in merger announcement stock returns seek to explain variation in either bidder and/or target returns in isolation (e.g. Hawanini and Swary, 1990; Houston and Ryngaert, 1992). A systematic analysis of cross-sectional differences in percentage revaluation of the combined bidder and target is absent in the literature. A notable exception is the recent study by Comett and Tehranian (1992). For a sample of thirty 1982-1987 mergers, they document that the merger announcement abnormal return to the combined bidder and target is correlated with the post-acquisition performance of the merged bank. While this does not inform us as to the characteristics of a profitable merger, it is reassuring in the sense that the stock market does seem to forecast to some degree how profitable the merging firms will be in the future.
3. Description
of sample
A sample of initial bank merger agreements announced during the period 1985-1991 was collected from three sources: a news search on Dow Jones News Retrieval, the bank mergers and acquisitions section of the Wall Street Journal (WSJ) Abstracts Index, and the mergers and acquisitions section of the New York Times (NYT) Abstracts Index. To remain in the sample, the merger agreement must be announced in the WSJ or NYT and stock return data must be available for both the bidding and target firms. To keep a relatively homogenous set of acquisitions, we exclude acquisition offers where the bidder is a thrift institution. 6 We also eliminate offers that cannot be legally completed within two years due to state banking laws. A disproportionate number of such mergers are never completed. Lastly, we eliminate “mergers of equals” from the sample because it is unclear which firm is the bidder. 7 We define a transaction as a merger of equals when either the assets or the equity value of the smaller firm would constitute over 45% of the combined assets or equity value of the two firms and the board of directors of the new firm will be composed of equal numbers of directors from each firm. Finally, we require that both the bidder and target have at least $100 million in assets.
6 This eliminated many New England savings banks from our sample. This also and loans that turned up in the search and holding companies for savings banks and ’ This is necessary when we consider separately the abnormal returns of the bidder and when we consider the relative performance of the bidding and target banks. exclusion of four merger announcements.
excluded savings thrifts. and target banks, This leads to the
1160
J.F. Houston, M.D. Ryngaert / Journal of Banking & Finance I8 (1994) 11X-11
76
The sample consists of 153 merger announcements that meet the above criteria. Of the announced mergers 131 were completed and 22 were not. Since the market may not know which deals will be completed at the time of the initial merger agreement, we analyze the completed and uncompleted merger agreements together and separately. For each deal, we attempt to obtain a copy of the initial merger agreement from 8K filings and proxy filings describing the merger terms. In many instances, we cannot get the full copy of the merger agreement. In these cases, we rely on descriptions of the deals reported in news stories or in the SEC filings of the bidders and targets. The above information sources are used to determine the announced mode of acquisition. Data on the market values for all of the bidders and targets are available using the CRSP master files. We also check the financial reports of the target and bidding firms to insure that the share numbers reported on CRSP are accurate.
4. Abnormal
returns to bank mergers
4.1. Constructing
abnormal
returns
To properly measure the returns to each merger we attempt to capture any “information leakage” pertaining to each deal. For each bidder and target we identify two dates: the agreement date and the leakage date. The agreement date for the bidder and target is the date of the initial merger agreement reported in the WSJ or NYT. The leakage date for the target is the first announcement that the target was a takeover candidate. We find this date by going back three months from the merger date on the Dow Jones News Wire, NYT index and WSJ index to identify any news story indicating that the company put itself up for sale, was in merger talks, or received a takeover proposal. If we find such a story we go back an additional three months to see if there was a prior story. If there are news stories going back more than six months we discard the observation from the sample. The leakage date for the bidder is the first date that the bidder reveals that it is in talks with the target firm or offers to acquire the target firm. Note that the target’s leakage date will never be after the bidder’s leakage date. Also, if there are no identifiable news leaks before the agreement date, then the leakage date is the agreement date. We use event study methodology to estimate the abnormal returns for the bidder and target. Using continuously compounded firm and market returns, market model parameters are estimated using the CRSP equal weighted market portfolio over day - 230 to day -31, where day 0 is the leakage date for the bidder or target. A firm’s abnormal returns are not included in the analysis if less than 75 days of returns are available in the pre-event period to estimate a market model. Abnormal returns for both the bidders and targets are calculated by
J. F. Houston, M.D. Ryngaert / Journal of Banking & Finance
18 (1994) 1155-11%
1161
summing up the abnormal returns based on the market model estimates for four days before the firm’s leakage date through the agreement date. This is referred to as a five-day abnormal return even though in some cases the return windows are much longer due to a lag between the leakage date and the agreement date. a Our measure of total merger-related revaluation for each merger (i = 1 to 153) is: VM@R,,
+ vh&M&
VI%&+ VA&,
(2)
where VM;; is the market value of the target firm’s stock five days before the first merger bid for the target, VM,, is the market value of bidding firm’s stock five days before its merger bid, and ARCi and AR,, are the abnormal returns for the ith target and bidder, respectively. We calculate this revaluation for each merger. Note that this measure gives the true percentage change in the value of the combined pre-merger firm. Because firms in our sample vary in the volatility of their underlying returns and because the windows for calculating abnormal returns for different deals vary in length, all our abnormal returns’ analyses employ weighted least squares means and regressions. The variance estimates used to perform WLS means and regressions are based on prediction error variances for the bidder and target abnormal returns and estimates of correlation between the market model error terms of the bidder and target firms. The details are given in the Appendix. 4.2. Estimates of abnormal returns Since the market may be unable to determine which mergers will be ultimately completed, we report the abnormal returns for all firms in our sample as one group. We also report separate results for deals that are completed and for deals that are canceled. This allows us to investigate whether there are any market-induced checks on the bank acquisition process. When mergers are initially announced, the market reaction is a snapshot summary of how favorably the market perceives the combination of the two firms. If the initial reaction is negative, this may compel management or shareholder groups to scuttle the agreement. If this is the case, we would expect more adverse market reactions to merger announcements that are later canceled. 9
* We also calculated announcement windows that begin only one day before the leakage date. Our primary results are essentially unchanged using the shorter windows. The primary drawback with the shorter window was that it appeared to miss some run-up in the target’s abnormal return. 9 Jennings and Mazzeo (1991) look at a similar issue for industrial mergers. They find no relationship between returns to bidders and whether a deal is ultimately canceled. They do not consider whether the total returns to a merger announcement affect the probabitity of the merger being consummated.
1162 Table 1 Abnormal
J.F. Houston, M.D. ~yngu~rt/Journal
of3anki~g & Finance 18 (1994) 11.55-1174
returns to bidders and targets for completed
Number of offers: Panel A: total abnormal returns Average five-day abnormal returns t-stat. (H,: mean equals zero) r-stat. (Ha: means of subgroups are equal) Pet. of deals with negative abnormal returns Panel 3: bidder abnormal returns Average five-day abnormal returns t-stat. (Ha: mean equals zero) t-stat. (Ha: means of subgroups are equal) Pet. of deals with negative abnormal returns Panel C: target abnormal returns Average five-day abnormal return t-stat. (Ho: mean equals zero) t-stat. (H,: means of subgroups are equal) Pet. of deals with negative abnormal returns
and canceled
bank merger agreements
All deals 153
Completed deals 131
Canceled deals 22
0.0038 1.17
0.0046 1.36
- 0.0043 0.43
49.0%
47.3%
59.1%
- 0.0232 -6.89 **
- 0.0225 -6.15 **
- 0.0293 -3.44 **
71.9%
71.0%
77.3%
0.1439 14.39 **
0.1477 13.75 **
0.0979 3.66 **
8.5%
8.4%
0.85
0.77
1.73
* indicates significance at the 0.05 level, * * indicates significance The t-statistics assume unequal variances for the two subgroups.
9.1%
at the 0.01 level.
Table 1 examines the abnormal returns to the bidding firm stock, the target firm stock, and the total abnormal returns. The average total abnormal return is 0.38% for all 153 deals and is 0.46% for the 131 completed deals. Neither return is significantly different from zero at the 10% level. Almost half of all completed deals have negative abnormal returns. Thus, there is little evidence that the average merger creates any value. Also, there is only weak evidence that the market reacts more negatively to subsequently canceled deals. While canceled deals have a negative total abnormal return of -0.43%, the difference between completed and canceled deals is not statistically significant. The insignificant total abnormal return is the result of a negative abnormal return to bidders offsetting a positive abno~al return to targets. Bidding firms realize, on average, a five-day abnormal return of -2.32%, which is statistically significant. Roughly 72% of the bidding firms realized a negative abnormal return. Bidding firms realized even lower abnormal returns for deals that were ultimately canceled, but again the returns were not significantly different than the returns realized from completed deals. As expected, target firms realize, on average, positive abnormal returns of 14.39% for the entire sample. The abnormal returns are considerably worse for the canceled deals (9.79% as compared to 14.77% for the completed deals). This difference is statistically significant at the 10% level. This suggests that target managements and/or shareholders may choose to back out of deals where merger premia are too small.
J.F. Houston, M.D. Ryngaert/Journal Table 2 Abnormal
returns by year, 19851991
ofBanking & Finance 18 (1994) 1155-1176
(all deals)
Variable
Five-day total abnormal returns
Five-day bidder abnormal returns
Five-day target abnormal returns
1985-1991 1985 1986 1987 1988 1989 1990 1991
0.004 0.005 0.005 - 0.019 * -0.011 0.006 0.017 0.033 *
- 0.023 ** - 0.025 ** 0.018 ** - 0.052 * * - 0.029 ** - 0.017 ** - 0.018 - 0.001
0.144 0.129 0.103 0.114 0.125 0.210 0.223 0.302
* indicates significance
1163
at the 0.05 level, ** indicates significance
** ** ** ** ** ** ** **
at the 0.01 level.
Table 2 summarizes the abnormal returns for each year in the sample. The results suggest that bank mergers have been viewed more favorably by the market in recent years. Prior to 1989, the average total abnormal return was negative. Since 1989, bank mergers have generated positive total abnormal returns, with the best year being the most recent, 1991, where total returns were a positive 3.3%. The improved total abnormal return results in the latter years of our sample is primarily reflected in higher target abnormal returns, although bidder abnormal returns have become less negative in recent years. In the subsequent section we demonstrate that some of this shift over time can be attributed to the changing characteristics of later mergers.
5. Cross-sectional
analysis
of abnormal
returns
While the average merger during our period of examination appears to yield only modest positive revaluations of the combined firms, there is considerable variation in those returns. To explain cross-sectional variation in merger announcement abnormal returns, we look at a number of factors including the operating performance of the bidder and target, the extent to which the operations of the target and bidder overlap, the financing of the deal, and the size of the deal. A traditional explanation for mergers in any industry is that the market for corporate control tends to move assets from lesser to better managed firms. This suggests that merger gains would tend to be larger when high performing banks acquire low performing banks. Similarly, one could argue that acquiring banks with a good track record tend to make better acquisitions. To address this issue we construct the following performance measures. For each bidder and target, we calculate the firm’s return on assets (net income before
1164
J.F. Houston, M.D. Ryngaert/Journal
ofBanking & Finance 18 (1994) 1155-1176
extraordinary items divided by average firm assets) less the median return on assets for all banking firms on the Compustat industrial and research tapes. This figure, NETROA, is measured for the fiscal year prior to the takeover bid announcement. To measure the performance of the bidder relative to the target, we construct ROADIFF, which is the difference between the bidder and the target return on assets in the year prior to the merger announcement. We expect the total return to a merger to be related to bidder NETROA and ROADIFF. The synergies created by a merger may also be related to the degree that two banking organizations have overlapping operations. So called in-market mergers have a greater potential to realize cost savings because of greater opportunities to consolidate back-office operations and close less efficient local branches. Published news reports frequently cite an ability to serve the same number of customers at a lower cost as one of the rationales for a merger. For instance, when Comerica announced its 1991 merger with Manufacturers National, the two firms reported that due to the considerable overlap of their operations in the Detroit area, the merged entity would close 60 branch offices in Michigan, reduce its work force by 13%, and cut annual costs by 15%. lo There is some evidence to suggest that in-market mergers are better received by the stock market. For example, Hawawini and Swary (1990) report that intrastate bank mergers result in positive stock price reactions and interstate bank mergers result in negative stock price reactions. By contrast, Cornett and De (1991) conclude that interstate bank mergers during the period 1983-1986 were value-increasing because both bidder and targets experience positive abnormal returns on the announcement of a merger. Unfortunately, defining a merger as interstate or intrastate may be an imprecise measure of the overlap between operations, and thus may provide little information concerning the potential for cost savings. For instance, two banks may operate in the same state, but they may operate in distinctly different parts of that state. Similarly, even though two banks may be headquartered in different states, the acquiring bank may have operations in the same markets as the target bank. To remedy this problem we construct a variable called OVERLAP. The variable is constructed as follows. First we identify each bank subsidiary of the target and bidding bank, including whenever possible deals that either bank may have pending. Using the Rand McNally and Thomson Bank Directory, we identify how many branch offices each firm has in a given city. ” Let n be the total number of cities that either bank has offices in and let Ti denote the total number
lo See the Wall Street Journal, October 29, 1991, p. A3. ‘I The Rand McNally Bank Directory became the Thomson list the banking offices of every bank subsidiary by city.
Bank Directory
in 1991. The directories
of offices the target has in city i and let Bi denote the total number of offices the
bidder has in city i. The variable OVERLAP is defined as: kmin(.l;, OvERLAp = !+---
Bi) (3)
CKW) i=l The variable can take on a m~imum value of .5 where there is complete overlap between the two firms’ operations and a minimum value of 0 where there is no overlap. For example, suppose that one bank has five offices in a city and the other bank has two offices. In this case the maximum number of overlapping offices would be two. So, we sum up the maximum number of overlapping offices and then deflate by the total number of offices of the two merging banks. In this sense, OVERLAP can be thought of as a crude measure of the percentage of offices of the combined firms that can be closed as a result of a merger. We expect higher levels of OVERLAP to correspond to higher total abnormal returns from the merger. We also control for the mode of acquisition to capture any information that the bidder may be conveying through the choice of financing. I2 For instance, the willingness to issue common stock on fixed terms may reveal that the bidder’s management believes that either the bidder’s stock or the potential benefits of the merger are overvalued. On the other hand, if the firm commits to issue equity, but the number of shares issued is made a function of the future price of the bidder’s stock, the bidder may be communicating that good news will soon be revealed about it (or the merit of the acquisition) before the deal is closed. Houston and Ryngaert (1992) refer to this as a conditional stock offer. l3 Similarly, a reliance on cash payment or issuance of preferred as opposed to common stock financing may be an attempt to communicate that the bidding firm’s stock is undervalued.
I2 While them are a large number of studies ~~rne~~~g that bidding Firms perform worse when they offer stock rather than cash in industrial mergers (see e.g., Travtos, 1987; Brown and Ryngaert, 19%) the evidence from bank mergers is mixed. Cornett and De (1991) find no significant relationship between bidder abnormal returns and the mode of acquisition. However, Hawawini and Swary f1990) find that bank bidders do receive higher abnormal returns when acquisition is paid for in cash. I3 For example, the bidder may agree to pay the target $80 worth of stock for each share tendered. If the bidder’s stock price at the time of the merger is $20 a share, the target will receive four shares of the bidder’s stock. However, if the bidders’ stock price falls to $16 a share, the target will receive five shares of stock. Thus, the target is protected against any adverse private information held by the bidder that may become public prior to the completion of the merger. This protection is potentially more valuable in bank mergers, since it often takes from six months to a year for a merger to be completed, given the need to obtain regulatory approval. See Houston and Ryngaert (1992) for a more complete description of the motivation for using conditional. stock offers in bank mergers, as well as empirical evidence concerning the frequency of such offers.
1985-1991
0.159 (0.126)
0.797 (0.359)
0.258 (0.420)
0.040 (0.160)
MKTRATIO = market value of equity (target)/ market value of equity (bidder + target)
Percent of deal financed with any type of stock
Percent of deal financed with conditional stock
Percent of deal financed with preferred stock
153
0.028 (0.056)
OVERLAP
Number of observations
0.0023 (0.0056)
-0.0011 (0.0057)
NETROA target
ROADIFF
0.0012 (0.0030)
Mean (std. dev.) 1985-1991
NETROA bidder
Variable
Table 3 Summary statistics of selected variables
30
0.074 (0.235)
0.223 (0.393)
0.690 (0.398)
0.187 (0.103)
0.021 (0.045)
0.0008 (0.0031)
0.0005 (0.0031)
0.0012 (0.0021)
Mean (std. dev.) 1985
29
0.064 (0.211)
0.344 (0.461)
0.852 (0.321)
0.180 (0.119)
0.025 (0.056)
0.0019 (0.0039~
-0.0013 (0.0044)
0.0006 (0.0022)
Mean (std. dev.) 1986
20
0.054 (0.167)
0.150 (0.366)
0.826 (0.372)
0.185 (0.123)
0.023 (0.033)
0.0021 (0.0043)
- 0.0008 (0.0041)
0.0013 (0.0035)
Mean (std. dev.) 1987
16
&
0.259 (0.415)
0.947 (0.145)
0.127 (0.118)
0.031 (0.051)
0.0020 (0.0108)
- 0.0027 (0.0112)
- 0.0006 (0.0039)
Mean (std. dev.) 1988
25
0.020 to.tooj
0.326 (0.457)
0.786 (0.380)
0.126 (0.124)
0.032 (0.070)
0.0025 (0.0063)
-0.0016 (0.0060)
0.0009 (0.0026)
Mean (std. dev.) 1989
10
(:I
0.200 (0.422)
0.500 (0.480)
0.156 (0.210)
0.011 (0.018)
0.0028 (0.0046)
- 0.0020 (0.0034)
0.0009 (0.0016)
Mean (std. dev.) 1990
23
0.018 (0.085)
0.241 (0.424)
0.880 (0.282)
0.132 (0.122)
0.048 CO.0791
0.0045 (0.0053>
- 0.0009 (0.0060)
0.0036 (0.0037)
Mean (std. dev.) 1991
J.F. Houston, M.D. Ryngaert /Journal
of Banking & Finance 18 (1994) 1155-1176
1167
Three variables are created to control for the mode of acquisition. We set the variable PCTSTOCK equal to the percentage of the target’s stock that is bought by issuing any kind of stock. We set PCTCOND equal to the percentage of any offer where common stock is to be issued, but the number of shares issued is a function of the bidder’s future stock price. I4 We set PREFER equal to the percentage of the offer that consists of preferred stock. We expect the coefficients on PCTSTOCK to be negative, PCTCOND to be positive, and PREFER to be positive. We also calculate a measure of the relative size of the target to the bidder. This variable, MKTRATIO, is the value of the target five days before the first announcement of a takeover bid divided by the value of the bidder and the target five days before the takeover bid. Table 3 presents some summary statistics concerning NETROA, ROADIFF, OVERLAP, MKTRATIO, and the financing variables. The statistics are broken down for each year in the sample. The average return on assets for the bidding banks for the entire sample is significantly above the industry median. Note, this measure reaches its highest level in 1991 which is also the year with the highest total abnormal returns. By contrast, the average return on assets for the target banks is significantly below the industry median. ROADIFF for the entire sample is significantly greater than zero. So, even though merger-related stock gains are near zero, assets are clearly flowing to better performing banks as a result of acquisition activity. The average degree of market overlap between bidders and targets as measured by OVERLAP is only 2.8%. This suggests that the number of combined offices of the merging firms that can be closed as a result of the merger is quite small. The degree of overlap has increased in recent years, with the notable exception of 1990 where there were only ten deals. The greater potential for cost savings resulting from increased market overlap may explain, in part, why abnormal returns have been higher in recent years. Table 3 also illustrates that stock is the primary financing vehicle, with the average percent of a deal financed with stock being 79.7%. Included in this category are conditional stock offers (25.8% of financing) and preferred stock (4% of financing). For the most part, there has been no pronounced shift over time in the method of financing. Prior to takeover activity, the target is 15.9% of the combined market value of the two firms. In recent years the market ratio of the target to the bidder has declined somewhat. Table 4 reports WLS regression results relating the total abnormal returns to various explanatory variables for the entire sample of announced deals. The regressions were run with and without separate year dummies that were included to capture any changes over time that were not captured in the explanatory variables.
‘4hat
conditional
stock offers will have positive values for both PCLSTOCK
and PCTCOND.
J.F. Houston, M.D. Ryngaert/Journal
ofBanking & Finance 18 (1994) 1155-1176
1169
The results indicate that the market responds more positively to mergers where the bidding firm has been performing well. Specifically, there is a positive relationship between the bidder’s NETROA and the five-day total abnormal return. This is significant at the 5% level for the specification that does not include year dummies, and significant at the 10% level for the specification that employs separate year dummies. The target’s NETROA is also positively related to the total abnormal return. This result is significant at the 10% level for the specification that includes the year dummies. We also interact the NETROA measures with the MKTRATIO variable. Arguably, this yields a more relevant performance measure, since we might expect that the total returns would be higher if a profitable bidder acquires a larger target. In fact, we find that there is a significantly positive relationship (at the 1% level) between the bidder’s interacted performance measure and the total abnormal return. There is a positive but insignificant relationship between the target’s interacted performance measure and the total abnormal return. To see if gains from a merger are larger when the target is a low return on asset firm and the bidder is a high return on asset firm, we use ROADIFF (bidder ROA minus target ROA) as an explanatory variable. This variable is insignificant, suggesting that any cross-sectional variation in gains in our sample is not attributable to moving the management of assets from less profitable to more profitable institutions. l5 Instead, the performance measures generally indicate that the revaluation of joint bidder and target value is higher when both the bidder and the target are performing better than the industry average. As expected the OVERLAP variable has a significant impact on the merger abnormal returns regardless of specification. This suggests that in-market mergers are perceived as more profitable than mergers involving banks that operate in different markets. The financing variables also influence abnormal returns as expected. The increased use of stock results in more negative returns, although the use of conditional or preferred stock results in more positive returns than using common stock with a fixed exchange ratio. The total abnormal return is also positively related to MKTRATIO. Finally, the year dummies do provide additional explanatory power when included in the regression specification. Consistent with the evidence presented in Table 2, the coefficients on the yearly dummies (not reported) indicate an upward trend in total merger returns since 1987. This suggests that increased merger returns in recent years cannot be explained solely in terms of higher bidder profitability or increased prevalence of in-market mergers.
” These results conflict somewhat with the results found by Hawawini and Swary (1990). They find a negative correlation between the bidding bank’s prior market performance and the abnormal returns following its acquisition. Our results are more consistent with Merck et al. (1990) who find, for nonbank mergers, a positive relationship between performance prior to a merger and the abnormal returns surrounding the merger’s announcement.
abnormal
ROA (target)
a t-statistics
are in parentheses;
Year dummies R2
0.2242 (4.36) * * NO 0.2587
-0.0218 t-O.651
0.0193 (1.07)
0.0481 (2.68) **
0.1737 (3.07) * * YES 0.2332
0.2027 (3.73) * * NO 0.1540 0.1967 (3.68) * * YES 0.3280
at the 0.01 level.
0.0698 (1.981* 0.0437 (1.37)
0.1729 (3.06) * YES 0.2321
0.2014 (3.70) ** NO 0.1538
0.0260 (1.43)
0.5455 (0.10) 0.0716 (2.04) *
0.0194 (1.08)
0.0258 (1.42)
0.0170 (1.89)
0.0257 (1.67) - 0.0180 (- 1.69)
0.6374 (0.12) 0.0448 (1.43)
0.0167 (1.881
0.0171 (1.90)
-0.3164 ( - 0.43)
- 0.0017 ( - 0.20) - 0.0150 (-1.49)
0.0278 (1.76) - 0.0183 (- 1.72)
0.0075 (0.211
-0.1442 (-0.19)
0.0167 (1.881
0.0173 (2.04) *
28.2454 (3.36) ** 7.2275 (1.36)
- 0.0011 (-0.131 - 0.0151 (-1.50)
0.0244 (1.69) - 0.0216 (-2.15) *
portfolio of the bidder and target as the dependent variable (completed
at the 0.05 level; ** indicates significance
0.1861 (3.34) ** YES 0.2682
* indicates significance
MKTRATIO = market value of equity (target)/market value of equity (bidder + target) OVERLAP
ROADIFF X MKTRATIO
ROADlFF = ROA (bidder)-
NETROA (target) X MKTRATIO
NETROA (bidder) X MKTRATIO
0.2123 (4.001** NO 0.2007
0.9099 (1.181
0.8706 (1,121
NETROA (target) = ROA (target) Median ROA (industry)
0.0586 (1.68)
2.6865 (1.83)
3.1400 (2.20) *
NETROA (bidder) = RCA (bidder) Median ROA (industry)
0.0317 (1.00)
0.0418 (2.35) *
0.0396 (2.11) *
0.0348 (1.881
Percent of deal financed with preferred stock
28.7477 (3.43) * * 7.6049 (1.42)
0.0169 (2.02) *
0.0189 (2.13) *
0.0187 (2.151*
Percent of deal financed with any type of stock
Intercept
0.0214 (1.351 - 0.0218 ( - 2.07) * 0.0026 (0.32) - 0.0182 (- 1.91)
returns of a value-weighted
Coefficient estimates a
using the five-day
Percent of deal financed with common stock
variables
regressions
- 0.0010 (-0.12) - 0.0193 (- 1.93)
Explanatory
Table 5 Cross-sectional deals)
J.F. Houston, M.D. Rynga~rf/Jo~nal
ofBanking & Finance I8 (1994) 1155-1176
1171
In Table 5, we present the results from the same set of regressions for the sample of deals that were ultimately completed. The results are very similar to those presented in Table 4. The financing variables remain significant and are of the expected sign. For the set of completed mergers, the bidder’s NETROA, and the bidder NETROA interacted with MKTRATIO are even stronger determinants of total abnormal returns. The coefficients on the OVERLAP measure also remain positive and highly significant. The MKTRATIO measure has considerably less explanatory power for the sample of completed mergers. Again, year dummies are significant when included. Table 6 reports WLS regressions run individu~ly on the bidder and target five-day abnormal stock returns. Interestingly, fewer of the performance variables remain significant in explaining target or bidder returns. This suggests that the division of rents may vary considerably across deals in a fashion unrelated to the bidder and target return on assets. Target abnormal returns appear to be marginally more affected by the performance of the bidder than by the performance of the target. Likewise, the target’s performance has a significantly positive influence on the bidder’s abnormal return in three of the four specifications. The measure of market OVERLAP has a positive effect on both the bidder and target abnormal returns, although the results suggest that the target bank generally receives a larger portion of the benefits resulting from the potential for future cost-cutting as a result of the merger. Given that the total return regressions tend to yield somewhat different inferences than the regressions looking at the bidder and target in isolation, it follows that it can be misleading to focus on bidder and targets in isolation. This is particularly true for the MKTRATIO variable. Bidders lose more on large acquisitions and targets receive less, yet when we regress total returns against this variable, the coefficient on MKTRATIO is positive and frequently significant. The reason is that the total revaluation resulting from larger deals puts more weight on the target return which is generally positive and less weight on the bidder return which is generally negative. l6 Thus, a more complete picture is obtained by looking at the overall gains (losses) from a sample of mergers rather than examining target and bidder returns separately.
‘“ievidence suggests that the relative size of the participants affects the division of gains (losses) from the merger. For our entire sample, bidders appear, on average, to overpay for targets. If bidders always overpay, for example, by 10% of the target firm’s value, bidders will suffer much larger share price declines on larger deals. Therefore, bidders may find it more difficult to justify large premiums for larger targets. If this is the case, the premiums (and overpayments) paid to larger targets will be smaller. Note that bidder abnormal returns can still be more negative for larger deals, because while the percentage overpayment to the target may be smaller, the overpayment is made to a larger target.
regressions
0.0068 (0.78)
0.0348 (1.83)
1.4255 (1.16)
Percent of deal financed with conditional stock
Percent of deal financed with preferred stock
NETROA (bidder) = ROA (bidder) - Median ROA (industry)
-0.0173
- 0.0031 (- 0.37)
0.0359 (1.92)
0.0063 (0.73)
- 0.0171 (- 1.77)
- 0.0037 ( - 0.45)
2.0445 (1.37)
0.0372 (1.81)
0.0055 (0.58)
- 0.0157 (-1.43)
- 0.0031 ( - 0.34)
0.0460 (2.33) *
0.0044 (0.48)
- 0.0155 (- 1.47)
- 0.0015 (-0.17)
8.3727 (1.86)
o.Gil12 (0.02)
0.0513 (1.86)
- 0.0689 (- 2.23) *
0.2032 (7.29) **
- 0.0280 ( - 0.47)
0.0439 (1.61)
- 0.0579 (- 1.86)
0.2089 (7.59) * *
7.8106 (1.32)
-0.0174 ( - 0.27)
0.0442 (1.50)
-0.0519 (- 1.54)
0.2109 (7.12) **
- 0.0379 ( - 0.61)
0.0394 (1.35)
- 0.0457 (- 1.37)
0.2150 (7.38) **
completed deals, N = 131
all deals, N = 153
all deals, N = 153
completed deals, N = 131
Dependent variable = five-day target abnormal return Coefficient estimates ’
Dependent variable = five-day bidder abnormal return Coefficient estimates a
of bidder and target returns
( - 1.77)
variables
Percent of deal financed with common stock
Intercept
Explanatory
Table 6 Cross-sectional
1174
J.F. Houston, M.D. Ryngaert/Journal
ofBanking & Finance 18 (1994)
1155-1176
6. Conclusions Our results uncover a number of findings concerning acquisitions of publicly traded banks during the period 19851991. First, there is no apparent positive revaluation of the combined bidder and target values at the announcement of these acquisitions. Positive returns to targets are canceled out by negative returns to bidders. This is consistent with accounting studies that find no apparent cost savings from the average large bank merger. We do find, however, that total returns have been higher in recent years. Second, bidding banks tend to be more profitable than target banks and more profitable than other banks in their industries. Third, the market responds most favorably to acquisition announcements by firms with a good past operating performance. Fourth, deals where there is a higher degree of market overlap, and therefore a greater potential for cost savings are viewed more positively by the stock market. Fifth, the return to bank mergers is related to the financing used by the acquiring bank. It is interesting to speculate why bank merger announcements do not generate more positive abnormal returns to a combined portfolio of bidders and targets. The most obvious explanation is that they create no real synergies. Alternatively, while there may be a fair number of “good” acquisitions with great opportunities for cost savings in our sample, the returns to these are offset by a significant number of ill advised acquisitions instigated by either overly optimistic or empire building management teams. Two other explanations that are more charitable to the role of bank mergers come to mind. First, the probability of a firm being a target may be better recognized than the probability of a firm being a bidder. Thus, target firm prices are bid up in advance of a merger bid, but bidder returns are not bid down. We have attempted to control for this possibility by measuring total returns from the date at which it is announced that a bank is a potential target. Second, since most takeover bids are stock financed, the acquisition announcement of a bidding bank may be sending out the signal that the firm’s stock is overvalued. Undervalued bidders may tend to refrain from making an acquisition. This suggests that estimated total returns may be a downward biased estimate of the synergies resulting from bank mergers.
Appendix The variance
of each bidder or target’s abnormal
return is given below.
1
- niR,,J2 VAR(CARi)
= s’
ni + $/N, I
+
‘=-“;i c (R,, I= - 230
-KJ2
J.F. Houston, M.D. ~~nguert/Jo~rn~l
of Banking & Finance 18 (1994) 1155-1176
1175
where 2
sample variance of market model error term for estimation period. = number of days in estimation of market model. = number of days in CAR estimation window. ‘i bi, ej = the beginning and end dates of the CAR estimation window. R mr = return to the market on date t. RI?5 = average return to market during estimation period. The variance of each merger i’s total abnormal return is given below, where: =
Zi
VAR(TCAR,)
= [V,i/(V,i
+ V~i)l”V~(CAR,i)
hi>] 2Vm(CARb,) + 2[yi/( V,i+ vbj>] V,t+ ‘bill Pbttnbi/nti) + [ Vbi/( V,i+
[‘hi/t
x [vAR(CAR,J
x VIW(CAR,J]‘~
TCAR, = total abnormal return to merger i. v,i = market value of the target in merger = market value of the bidder in merger vbi = estimated correlation between bidder Pbt for estimation of marked model prior = number of days in bidder’s abnormal nbi = number of days in target’s abnormal nti
i five days before leakage date. i five days before leakage date. and target market model residuals to the target’s leakage date. return window. return window.
References Berger, A. and D. Humphrey, 1992, Megamergers in banking and the use of cost efficiency as an antitrust defense, The Antitrust Bulletin 33, forthcoming. Boyd, J. and S. Graham, 1991, Investigating the Bank Consolidation Trend, Federal Reserve Bank of Minneapolis Quarterly Review 1.5, 3-15. Bradley, M., A. Desai and E.H. Kim, 1988, Synergistic gains from corporate acquisitions and their division between the sharehoIders of target and acquiring firms, Journal of Financial Economics 21, 3-40. Brown, D. and M. Ryngaert, 1991, The mode of acquisition in takeovers: taxes and asymmetric information, Journal of Finance 46, 653-669. Comett, M.M. and S. De, 1991, Common stock returns to corporate takeover bids: evidence from interstate bank mergers, Journal of Banking and Finance, 273-296. Cornett, M.M. and H. Tehranian, 1992, Changes in corporate performance associated with bank acquisitions, Journal of Financial Economics 31, 211-234. Desai, A. and R. Stover, 1985, Bank holding company acquisitions, stockholder returns and regulatory uncertainty, Journal of Financial Research 8, 145-156. Gorton, G. and R. Rosen, 1992, Overcapacity and exit from banking, University of Pennsylvania working paper. Hawawini, G. and I. Swary, 1990, Mergers and acquisitions in the U.S. banking industry (Elsevier Science Publishers, New York).
Houston, J. and M. Ryngaert, 1992, Resolving information asymmetries in bank acquisitions: the use of conditional stock offers, University of Florida working paper. James, C. and P. Weir, 1987, Returns to acquirers and competition in the acquisition market: the case of banking, Journal of Political Economy 95, 355-370. Jennings, R. and M. Mazzeo, 1991, Stock price movements around acquisition announcements and management’s response, Journal of Business 64, 139-163. Mitchell, M. and K. Lehn, 1990, Do bad bidders become good targets? Journal of Political Economy 98, 372-398. Merck, R., A. Shleifer, and R. Vishny, 1990, Do managerial objectives drive bad acquisitions? Journal of Finance 45; 31-48. Neely, W., 1987, Banking acquisitions: acquirer and target shareholder returns, Financial Management 16,66-73. Spindt, P. and V. Tahran, 1992, Are there synergies in bank mergers? Presented at the 1992 Western Finance Association meetings. Srinivasan, A., 1992, Are there cost savings from bank mergers? Federal Reserve Bank of Atlanta Economic Review (March/April), 17-28. Srinivasan, A. and L. Wall, 1992, Cost savings associated with bank mergers, Federal Reserve Bank of Atlanta Working Paper 92-2. Stulz, R., R. Walkling and M. Song, 1990, The distribution of target ownership and the division of gains in successful takeovers, Journal of Finance 45, 817-834. Travlos, N., 1987, Corporate takeover bids, the method of payment and bidding firms’ stock returns, Journal of Finance 42, 943-963.