Cluster analysis of the financial characteristics of depository institution merger participants and the resulting wealth effects.

Lacewell, Stephen K. ; White, Larry R. ; Young, Michael T. 等

INTRODUCTION

The merger and acquisition activity of depository institutions has increased dramatically in recent years, with various theories hypothesized regarding the cause of this action. In addition, much attention has been devoted to the merger and acquisition activities of non-financial firms, with most utilizing a cross-sectional regression analysis to explain the pre- and post-merger returns of both bidders and targets. While this method has met with much success, it does suffer from some apparent weaknesses in its explanatory power. Alternative statistical techniques, many of which are found in the non-parametric realm of tests, have been implemented in attempts to offset the drawbacks of ordinary least squares regression. In this paper cluster analysis is used to analyze the mergers of depository institutions, drawing on previous work in the non-financial sector. Mergers are separated by pre-merger financial characteristics of both bidders and targets. This insures that the within-group differences are small relative to among-group differences. This provides a homogeneous group of mergers, which is analyzed for among-group differences in returns to both parties, allowing the financial data for both bidding firms and target firms to capture potential interactions among their characteristics.

As noted earlier, the merger and acquisition activity of depository institutions can be described as at least frenzied during the decade of the 90's. Researchers attempt to explain this phenomenon through deregulation, competition, institutional efficiency, market share, product diversity and a host of other reasons. Efforts are also made to utilize non-financial institution merger hypotheses to explain the cause and effect of this activity, with results that can be described as contradictory at best. However, considering basic differences in financial make-up and the merger process in general, there are various reasons to expect differences in merger studies targeting depository financial institutions versus non-financial institutions. As stated by Chang, Gup and Wall (1989) depository institution mergers take an extensive amount of time (which increases merger uncertainty) due to the need for regulatory approval. Also, the nature of the assets acquired differs significantly when comparing depository institutions and industrial firms. When a bidder successfully acquires a depository institution target, it is buying a set of relationships generated by the existing management rather than a set of physical assets [Baradwaj et al. (1990)]. The uncertainty related to keeping current management in place makes this type of merger fundamentally different than those of other industries. Additional differences between the industrial and bank sector addressed by Zhang (1998) include degree of regulation, degree of competition and ownership structure. Degree of regulation refers to the fact that banks exert disproportional influence on the economy as compared to non-financial firms. Thus, for reasons of financial prudence and monetary control, the bank sector and mergers between its participants are heavily regulated. Degree of competition simply refers to non-bank financial institutions' increasing presence in areas traditionally dominated by banks. As for ownership structure, bank takeovers differ in that they do not produce wealth effects on the bondholders (depositors), who receive explicit and implicit protection from the Federal Deposit Insurance Corporation. Thus, the aforementioned diverse results may be due to the various motivations of bidding firms, financial (depository) versus non-financial firm differences, or more likely, a combination of both. The majority of previous studies focus on single motivation factors or limited participant characteristics and fail to account for the heterogeneity within merger samples. This paper furthers the merger literature by separating depository institution mergers into homogeneous groups of bidders and targets based on the pre-merger financial characteristics of each. This allows the analysis of the influence of intergroup differences on the returns to both bidding and target firms.

LITERATURE REVIEW

This study is based largely on the work of Sawyer and Shrieves (1994) who analyze non-financial firm mergers between 1975 and 1987. The authors utilize the cluster analysis approach for merger differentiation to avoid weaknesses of past studies that involve the use of cross-sectional regression of merger returns on variables theorized to affect the profitability of an acquisition or tender offer to either or both firms involved. As noted by Sawyer and Shrieves potential problems with the regression approach include (1) theories other than those being tested are relevant to the sample of acquisition events used; (2) the explanatory variables have alternative interpretations; (3) the explanatory variables interact in complex and unanticipated ways; (4) the explanatory variables have nonlinear relations to merger returns; and (5) the variables included are relevant only in a subset of the observations. Thus, attempts are made to utilize their methodology, and its attractive features, to gain insight into the merger motivation of depository institutions.

Studies related to depository institution mergers and common stock returns are many and diverse, although the attention devoted to this topic pales in comparison to merger studies regarding industrial firms. Previous works tend to focus on specific aspects of a merger, such as intrastate versus interstate, large versus small institutions, how closely held is the target versus the bidder, etc. Due to the variety of studies available, a brief survey of financial institution literature is presented to illustrate the inconclusiveness of results to date.

Among the earliest studies is Cornett and De's (1991) investigation of the stock market's reaction to the announcement of interstate bank mergers. In contrast to studies of non-financial mergers, the research finds significant positive announcement-period excess returns for both bidding and target banks. As for intrastate bank mergers, James and Weir (1983) and Desai and Stover (1985) also document positive announcement-period bidder returns.

Baradwaj, Fraser and Furtado (1990) examine hostile bank takeovers and find hostile bank acquisition announcements produce positive net wealth effects which are larger than the wealth effects of nonhostile acquisitions. Whalen's (1997) event study analysis of intracompany bank mergers reveals significant, positive average and cumulative average abnormal returns following a merger. Subrahmanyam, Rangan and Rosenstein (1997) report a negative relation between abnormal returns and the proportion of independent outside directors on the board of bidding banks.

While the majority of merger studies focus on independent gains to bidders and targets, several papers examine the consolidated abnormal returns to mergers. Hannan and Wolken (1989), Houston and Ryngaert (1994), and Pilloff (1996) find small or no average wealth creation resulting from bank mergers. These findings agree with a study by Madura and Wiant (1994). Their analysis of 152 mergers between 1983 and 1987 find negative cumulative abnormal returns for acquirers during the 36-month period following the merger announcement. However, Cornett and Tehranian (1992) not only find average merger-related gains among their sample of thirty mergers involving publicly traded banking institutions, but also conclude that weighted abnormal returns around the merger announcement are positive. These findings are consistent with the analysis by Zhang (1995) of 107 mergers taking place between 1980 and 1990, which finds a significant increase in overall value resulting from the mergers. Furthermore, Gupta, LeCompte and Misra (1997), in an examination of solvent-stock-held savings institutions from the late 1970s to the early 1990s, show losses to acquiring firms, significant earnings of target stockholders and positive wealth effects to the bidder-target pair.

The mixed results achieved by researchers point to weaknesses in current methodology. The analysis presented here attempts to expand the understanding of depository institution mergers through the use of both parametric and non-parametric methods, as discussed in the following section.

DATA AND METHODOLOGY

The initial merger sample is obtained from the Sheshunoff BankSearch Mergers and Acquisitions database utilizing completed mergers with an announcement date between January 1, 1993 and September 30, 1999. Additionally, comprehensive financial information on both bidder and target and the method of payment to target is obtained from Sheshunoff. To allow for comparability, only cash-for-stock (cash) and stock-for-stock (stock) transactions are retained in the sample. The initial sample of mergers announced and completed between 1993 and 1999 consisted of 2,967 mergers. However, many depository institutions, especially small targets, are not actively traded on an organized exchange. Thus, the number of firms that satisfy the criteria of having continuous daily transactions was 212 Bidders and 190 Targets. Daily returns for bidders and targets are then obtained from the Center for Research in Security Prices (CRSP) database to be used in merger return calculations.

Cluster Analysis

To successfully conduct the cluster analysis, several variables are needed for each of the participants in the merger transaction to form the cluster variate. The following variables are obtained for both bidders and targets over a three-year period beginning with the most recent year-end prior to the merger announcement date unless otherwise noted:

* Natural log of total assets pre-merger;

* Return on assets;

* Return on equity

* Ratio of equity to total assets;

* Ratio of core deposits to total deposits;

* Ratio of non-performing assets to total assets;

* Ratio of operating expense to average assets;

* Efficiency ratio;

* Ratio of noninterest income to average assets;

* Price to market price one year prior;

* Number of branches;

The natural log of total assets pre-merger, or firm size, of bidders and targets has a number of interpretations. A large bidder and a small target is the most common type of depository institution merger. If one party is small relative to its minimum efficient size, economies of scale may be presented as a motivating factor. A large bidder wishing to maximize geographic coverage or market share may merge with a large target. The number of branches indicates future growth opportunities or areas of cost savings if overlapping branches are closed.

The pre-merger return on assets and return on equity are profitability measures used to separate firms based on positive or negative earnings trends, and the implications of each denote positive or negative connotations. The ratio of noninterest income to average assets is used to show a firms revenue generation by activities outside the traditional loan and investment portfolio. Targets with above average noninterest income would tend to be very attractive to bidders wishing to expand income sources.

Equity to assets is a measure of capital adequacy, both to shareholders and regulators. This is a reflection of past managerial performance and future growth opportunities. Core deposits to total deposits provides a measure of funds stability, which is desirable in both bidders and targets. Non-performing assets to total assets is included to assess asset quality, which reflects both managerial performance and local economic conditions.

The ratio of operating expenses and the efficiency ratio provide input as to how well bidders and targets manage overhead expenses. The efficiency ratio is defined as the last twelve months noninterest expense divided by interest income plus noninterest income. An above average ratio for either of these variables points to possible areas of future cost reductions, which should lead to greater efficiency for the bidding firm. Finally, the number of branches is used as a measure of relative size and market coverage for both bidders and targets.

As per Hair (1995) the data are standardized using a general form of a normalized distance function, which utilizes a Euclidian distance measure amenable to a normalizing transformation of the raw data. This process converts the data into a standard normal value with a zero mean and a unit standard deviation. This transformation, in turn, eliminates the bias introduced by differences in scales of the variables used in the analysis.

Merger Returns Methodology

The methodology used to assess statistical significance of merger returns begins with the market model as follows:

[R.sub.i,t] = [a.sub.i] + [B.sub.i] [R.sub.m,t] + [e.sub.i,t]

where [R.sub.i,t] is the daily return on firm i's stock in period t and [R.sub.m,t] is the return on the value-weighted CRSP index in period t. Parameters ai and Bi are estimated over the base period t = -270 to t = -21, with t = 0 the announcement date (AD) of the merger. The abnormal return for each firm, [AR.sub.i,t], is calculated for the period t = -20 to t = +20 and t = -1 to t = 1 and is given by:

[AR.sub.i,t] = [R.sub.i,t] - ([a.sub.i] + [B.sub.i] [R.sub.m,t])

Average abnormal returns, [AAR.sub.t], for the N firms for each day (t) is calculated as:

[AAR.sub.t] = Sum [AR.sub.i,t] x (1/N)

and the cumulative abnormal return, CAR, for any period T is calculated as:

[CAR.sub.T] = Sum [AAR.sub.t]

A binary variable, STOCK, is used to control for other factors that may affect the returns of mergers involved in the study. The use of this variable allows the analysis of the type of consideration used in the merger, and equals one for stock transactions and zero for cash transactions. Also, as documented by Sawyer and Shrieves in their original study, we must consider the interaction between type of consideration and cluster membership to determine whether the clusters of mergers have differing effects on returns depending on consideration type.

The variable RELSZ is used to determine if there is an interaction effect between the type of consideration and relative bank size. It is calculated as the natural logarithm of the ratio of total assets of the target to total assets of the bidder. Sawyer and Shrieves note a working paper by Asquith, Bruner and Mullins (1987) that finds a statistically significant negative relation between bidding firm returns and the relative size of the target in stock transactions, and an insignificant, although positive, relation when cash is the type of consideration utilized.

Considering the aforementioned information the following model is estimated to determine merger returns:

[CAR.sub.j] = [a.sub.0] + [a.sub.1][RELSZ.sub.j] + [a.sub.2][STOCK.sub.j] + [a.sub.3]([STOCK.sub.j] x [RELSZ.sub.j]) + Sum [b.sub.i][CLSTR.sub.ij] + Sum [c.sub.i]([CLSTR.sub.ij] x [STOCK.sub.ij]) + [e.sub.j]

where:

[CAR.sub.j] = cumulative abnormal return

[a.sub.0] = intercept meant to capture the mean effect of CAR for cash transactions in the reference cluster, i.e., for observations where [CLUSTR.sub.ij] = 0 for i = 1, ..., N-1

[a.sub.0] + [a.sub.2] = the estimate of the effect of stock transactions on CAR in the reference cluster where [STOCK.sub.j] = 1 for such transactions

[a.sub.1] = effect of size on cash transactions

[a.sub.1] + [a.sub.3] = effect of size on stock transactions

[b.sub.i] = estimate for the differential effect of cluster i on CAR (relative to the reference cluster) for cash transactions, where i = 1, ..., N-1

[b.sub.i] + [c.sub.i] = estimate for the differential effect of membership in cluster i on stock transactions

Hypothesis tests for significance relating to the effect of clusters on CARs are:

Cluster effects may be present in cash deals;

[H.sub.0]: [b.sub.i] = 0 for i = 1, ..., N-1 (equality of cluster means in cash deals)

[H.sub.1]: One or more [b.sub.i] [not equal to] 0

Cluster effects may be present in stock deals; [H.sub.0]: [b.sub.i] + [c.sub.i] = 0 for i = 1, ..., N-1 (equality of cluster means in stock deals) [H.sub.1]: One or more bi + ci [not equal to] 0

Cluster effects may be present across either cash or stock transactions; [H.sub.0]: [b.sub.i] = [c.sub.i] = 0 for i = 1, ..., N-1 (equality of cluster means in cash or stock deals) [H.sub.1]: One or more bi or ci [not equal to] 0

Hypothesis tests for significance relating to the effect of relative size on three-day CARs are:

Size has no effect on CARs; [H.sub.0]: [a.sub.1] + [a.sub.3] = 0

[H.sub.1]: [a.sub.1] + [a.sub.3] [not equal to] 0

Hypothesis tests for significance relating to the effect of type of consideration on three-day CARs are:

Type of consideration has no effect on CARs; [H.sub.0]: [a.sub.3] = [a.sub.2] = [c.sub.1] = [c.sub.2] = ... = [c.sub.N-1] [H.sub.1]: One or more of the coefficients [not equal] 0

RESULTS

The cluster analysis of financial characteristics results in six distinct clusters, although only four are of usable size for the analysis. The financial highlights of these four clusters are given below with the variable means summarized in Exhibit 3.

Cluster Results

Cluster 1 (Contains 27 mergers)

There is a considerable size difference between bidders, with average assets of slightly more than $4 billion, and the targets, with average assets of slightly more than $400 million. Additionally, the average number of branches for bidders, at 177, is almost six times the number of branches (30) of the average target. Bidders exhibited greater profitability than targets; however, both bidders and targets in this cluster were relatively profitable, by traditional measures. ROAs were 1.26 percent and 1.08 percent, for bidders and targets, respectively. Bidders' ROE was 15.58 percent, while the average ROE for targets was 11.82 percent. Additionally, both bidders and targets enjoyed a relative advantage in their efficiency ratio in comparison to members of the other clusters. Overall, the financial characteristics of Cluster 1 bidders and targets would tend to support the value-maximizing hypothesis of merger activity.

Cluster 3 (Contains 28 mergers)

This cluster is characterized by the most significant size differences between bidders, with average assets of more than $51 billion, and targets, with average assets of approximately $1.25 billion. Additionally, the bidders operated over 10 times as many branches as the average target. In this cluster, the bidders exhibited a much greater profitability than the targets. The average bidder had an ROA of 1.37 percent compared to the average target's ROA of 0.95 percent. Bidders' average ROE were over 18 percent compared to the targets' average ROE of approximately 12 percent. Both bidders and targets were relatively inefficient; however, the bidders and targets both enjoyed good non-interest income support from their operations. Overall, the financial characteristics of Cluster 3 bidders and targets would tend to support the size-maximizing hypothesis.

Cluster 5 (Contains 68 mergers)

There is a striking similarity in the financial characteristics of the bidders and targets in this cluster. The size difference between bidders and targets, with the average assets of the bidders at approximately $15 billion and the targets at just over $1 billion, is the major characteristic difference between the two groups. The other ratios are very similar between the bidders and targets, leading us to propose that these mergers tend to support the size-maximizing hypothesis.

Cluster 6 (Contains 86 mergers)

The size difference between bidders and targets is the smallest in this cluster, as measured by average assets ($4.2 billion for bidders, $442 million for targets) and number of branches (118 for bidders, 16 for targets). The bidders tend to be significantly more profitable, with average ROAs of 1.24 percent versus 0.95 percent and average ROEs of 14.86 percent versus 11.02 percent. The bidders also were more efficient and enjoyed a greater contribution of non-interest income than their targets, on average. Overall, the financial characteristics of this cluster would tend to support the improved-management hypothesis.

Return Results

In the bidder regression (Exhibit 1), relative size between the bidders and targets, as well as the form of compensation for the merger/acquisition, specifically stock transactions, are significant determinants of cumulative abnormal returns. These findings are in line with other studies and are as expected. In the target regression (Exhibit 2), there are no statistically significant determinants of cumulative abnormal returns, nor is the model statistically significant. While this outcome is surprising, it is posited that model misspecification, as well as poor variable selection for the cluster analysis, may be underlying causes. However, it should be noted that this finding is fairly consistent with the results of Sawyer and Shrieves.

Overall, bidders experienced a statistically significant positive abnormal return the day before the announcement of the merger and a statistically significant negative abnormal return on the day of the merger announcement, which lasted through the third day following the merger announcement. The bidders' negative cumulative abnormal return began over two weeks in advance of the merger announcement, with the major negative movement occurring at the announcement day and continuing well beyond the end of the evaluation period. Clusters 1 and 5 experience extremely negative cumulative abnormal returns resulting in the decline of shareholder wealth through day +20 of the merger announcement, indicating that possibly these buyers were being punished by the market for a poor purchase choice (Exhibit 4). The findings related to buyers are consistent with previous merger studies.

Shareholders of targets, as expected, are the relative beneficiaries of the resulting wealth effects, experiencing positive abnormal returns at the announcement of the merger. Their positive cumulative abnormal returns commenced with the merger announcement and continue well beyond the end of the evaluation period (Exhibit 6). When evaluating the individual bidder and target cluster abnormal and cumulative abnormal returns, the relationships that are observed from the combined data analysis are still present; however, it is interesting to observe the relative differences between the four clusters as shown in the graph. Shareholders of banks in clusters 2, 5, and 6 are especially benefitted by their mergers. Overall, the reaction of the market in rewarding the targets' shareholders is consistent with previous merger findings.

[GRAPHIC OMITTED]

[GRAPHIC OMITTED]

[GRAPHIC OMITTED]

CONCLUSIONS

The merger and acquisition activity of depository institutions increased dramatically during the 1990s, with various theories hypothesized regarding the cause of this marked increase. Previous studies, generally, have focused on single factors based on the researchers' preconceived ideas regarding the motivation for mergers. Given the number of merger theories, studies based on one motivating factor may not be able to identify results that validate an alternative. To avoid this problem, this paper separates depository institution mergers into homogeneous groups of bidders and targets based on the pre-merger financial characteristics of each by using the non-biased method of cluster analysis. The resulting wealth effects accruing to both buyers and targets are as expected, buyers generally lose and targets gain, and are consistent with previous studies of this type. The magnitude of the gains and losses vary from cluster to cluster indicating that there are combinations of characteristics which result in "better" or "worse" mergers. However, it is important to note that this paper is a work in progress. While some of findings support existing merger theories, others are marginal at best. The investigation will continue with added results forthcoming.

REFERENCES

Asquith, P., R.F. Bruner &D.W. Mullins. (1987). Merger returns and the form of financing. Unpublished manuscript.

Baradwaj, B.G., D.R. Fraser & P.H. Furtado. (1990). Hostile bank takeover offers. Journal of Banking and Finance, 14, 1229-1242.

Berger, A.N., W.C. Hunter & S.G. Timme. (1993). The efficiency of financial institutions: A review and preview of research past, present, and future. Journal of Banking and Finance, 17, 221-249.

Brown, S.J. & J.B. Warner. (1985). Using daily stock returns: The case of event studies. Journal of Financial Economics, 14, 3-31.

Chang, D., B. Gup & L. Wall. (1989). Financial determinants of bank takeovers. Journal of Money, Credit, and Banking, 524-536.

Cornett, M.M. & S. De (1991). Common stock returns in corporate takeover bids: Evidence from interstate bank mergers. Journal of Banking and Finance, 15, 273-295.

Cornett, M.M. & H. Tehranian. (1992). Changes in corporate performance associated with bank acquisitions. Journal of Financial Economics, 31, 211-234.

Curry, T.J. & J.T. Rose. (1997). Thrift strategies after FIRREA: A cluster analysis of high-performance institutions. Journal of Economics and Business, 49, 223-238.

Desai, A. & R. Stover. (1985). Bank holding company acquisitions, stockholder returns and regulatory uncertainty. Journal of Financial Research, 8, 145-156.

Gupta, A., R. LeCompte & L. Misra. (1997). Acquisitions of solvent thrifts: Wealth effects and managerial motivations. Journal of Banking and Finance, 21, 1431-1451.

Hair, J.F., R.E. Anderson, R.L. Tatham & W.C. Black (1995). Cluster Analysis, Multivariate Data Analysis. Englewood Cliffs, NJ: Prentice Hall.

Hannan, T.H. & J.D. Wolken. (1989). Returns to bidders and targets in the acquisition process: Evidence from the banking industry. Journal of Financial Services Research, 3, 5-16.

Houston, J.F. & M.D. Ryngaert (1994). The overall gains from large bank mergers. Journal of Banking and Finance, 18, 1155-1176.

James, C. & P. Weir (1983). Returns to acquirers and competition in the acquisitions market: The case of banking. Journal of Political Economy, 95, 355-370.

Lee, D. (1998). The gains from takeover deregulation: Evidence from the end of interstate banking restrictions. Journal of Finance, 53, 2185-3001.

Madura, J. & K.J. Wiant. (1994). Long-term valuation effects of bank acquisitions. Journal of Banking and Finance, 18, 1135-1154.

Malatesta, P. & R. Thompson. (1985). Partially anticipated events: A model of stock price reactions with an application to corporate acquisitions. Journal of Financial Economics, 14, 237-250.

Pettway, R.H. & T. Yamada. (1986). Mergers in Japan and their impacts upon shareholders' wealth. Financial Management, 15, 43-52.

Pilloff, S.J. (1996). Performance changes and shareholder wealth creation associated with mergers of publicly traded banking institutions. Journal of Money, Credit and Banking, 28, 294-311.

Sawyer, G.M. & R.E. Shrieves. (1994). Stockholder returns among homogeneous groups of mergers. Journal of Financial Research, 17, 45-63.

Scanlon, K.P., J.W. Trifts & R.H. Pettway. (1989). Impacts of relative size and industrial relatedness on returns to shareholders of acquiring firms. Journal of Financial Research, 12, 103-112.

Subrahmanyam, V., N. Rangan & S. Rosenstein. (1997). The role of outside directors in bank acquisitions. Financial Management, 26, 23-37.

Trifts, J. & K. Scanlon. (1987). Interstate bank mergers: The early evidence. Journal of Financial Research, 10, 305-311.

Whalen, G. (1997). Wealth effects of intraholding company bank mergers: Evidence from shareholder returns. Managerial Finance, 23, 91-107.

Zhang, H. (1995). Wealth effects of U.S. Bank Takeovers. Applied Financial Economics, 5, 329-336.

Zhang, H. (1998). U.S. evidence on bank takeover motives: A note. Journal of Business Finance and Accounting, 25, 1025-1032.

Stephen K. Lacewell, Murray State University

Larry R. White, Mississippi State University

Michael T. Young, Minnesota State University

Exhibit 1
Buyer Regression Data:

[CAR.sub.j] = [a.sub.0] + [a.sub.1][RELSZ.sub.j] + [a.sub.2]
[STOCK.sub.j] + [a.sub.3]([STOCK.sub.j] x [RELSZ.sub.j]) + Sum
[b.sub.I][CLSTR.sub.ij] + Sum [c.sub.i]([CLSTR.sub.ij] x
[STOCK.sub.ij]) + [e.sub.j]

 Bidders

 N 212
 Adjusted [R.sup.2] 0.0921
 Overall F-Statistic 3.62
 Probability Level 0.0006

Variable Coefficient t-value P > t

INTERCEPT [a.sub.0] -0.02043 -1.03 0.3048
RELSZ [a.sub.1] -0.00865 -1.43 0.1553
STOCK [a.sub.2] -0.00610 -0.29 0.7689
STOCK*RELSZ [a.sub.3] 0.00166 0.26 0.7955
CLSTR1 [b.sub.1] 0.00575 0.74 0.4587
CLSTR3 [b.sub.2] -0.00478 -0.27 0.7877
CLSTR5 [b.sub.3] -0.03513 -2.08 0.0392
CLSTR*STOCK3 [c.sub.1] 0.00092 0.05 0.9631
CLSTR*STOCK5 [c.sub.2] 0.01977 1.10 0.2730

TESTS:

 F value Pr > F

 [b.sub.1] = 0: 0.55 0.4587
 [b.sub.2] = 0: 0.07 0.7877
 [b.sub.3] = 0: 4.31 0.0392
 [b.sub.2] + [c.sub.1] = 0: 0.17 0.6781
 [b.sub.3] + [c.sub.2] = 0: 6.37 0.0124
 [b.sub.2] = [c.sub.1] = 0: 0.12 0.8846
 [b.sub.3] = [c.sub.2] = 0: 5.34 0.0055
 [a.sub.1] + [a.sub.3] = 0: 11.57 0.0008
 [a.sub.1] = [a.sub.3] =
 [c.sub.1] = [c.sub.2] = 0: 0.46 0.7655

NOTE: Cluster number one contained only stock purchases. Therefore,
the last term in the above model sums to N-2 rather than N-1 as would
normally be the case. Cluster number six was used for the reference
cluster in the regression and clusters two and four were omitted from
the regression due to their small size.

Exhibit 2

Target Regression Data

[CAR.sub.j] = [a.sub.0] + [a.sub.1][RELSZ.sub.j] + [a.sub.2]
[STOCK.sub.j] + [a.sub.3]([STOCK.sub.j] x [RELSZ.sub.j]) + Sum
[b.sub.I][CLSTR.sub.ij] + Sum [c.sub.i]([CLSTR.sub.ij] x
[STOCK.sub.ij]) + [e.sub.j]

 Targets

 N 190
 Adjusted [R.sup.2] 0.0075
 Overall F-Statistic 1.15
 Probability Level 0.3276

 Variable Coefficient t-value P > t

INTERCEPT [a.sub.0] 0.23561 1.54 0.1257
 RELSZ [a.sub.1] -0.01666 -0.52 0.6062
 STOCK [a.sub.2] -0.07721 -0.49 0.6232
 STRELSZ [a.sub.3] 0.00479 0.14 0.8883
 CLSTR4 [b.sub.1] -0.19317 -1.95 0.0532
 CLSTR5 [b.sub.2] -0.21611 -2.17 0.0312
 CLSTR6 [b.sub.3] -0.14609 -1.04 0.2999
 CLSTOCK4 [c.sub.1] 0.16848 1.55 0.1240
 CLSTOCK5 [c.sub.2] 0.20623 1.99 0.0481
 CLSTOCK6 [c.sub.3] 0.09863 0.66 0.5120

TESTS:

 F value Pr > F

 [b.sub.1] = 0: 3.79 0.0532
 [b.sub.2] = 0: 4.72 0.0312
 [b.sub.3] = 0: 1.08 0.2999
 [b.sub.1] + [c.sub.1] = 0: 0.30 0.5847
 [b.sub.2] + [c.sub.2] = 0: 0.12 0.7332
 [b.sub.3] + [c.sub.3] = 0: 0.81 0.3702
 [b.sub.1] = [c.sub.1] = 0: 2.04 0.1325
 [b.sub.2] = [c.sub.2] = 0: 2.42 0.0922
 [b.sub.3] = [c.sub.3] = 0: 0.94 0.391
 [a.sub.1] + [a.sub.3] = 0: 1.19 0.2764
 a = a = c = c = c = 0: 1.43 0.214

NOTE: Cluster number two was used for the reference cluster in the
regression and clusters one and three were omitted from the
regression due to their small size.

Exhibit 3: Resulting Clusters Based on Financial Characteristics

CLUSTER # 1 2 3
FREQ 27 1 28
BUYER
BTA (000) 4,073,000 28,346,099 51,136,035
BNLTA 15.22 17.16 17.75
BROA 1.26 15.26 1.37
BROE 15.58 14.51 18.60
BEA 8.60 10.88 7.33
BCDTD 0.00 0.00 0.86
BNPAA 0.60 NA 1.37
BOEAA 2.95 25.24 3.59
BER 58.07 43.38 62.51
BNIIAA 1.24 21.65 1.67
BPPL 146.54 200.05 196.69
BNB 176.62 508.00 1029.74
TARGET
TTA (000) 404,335 408,399 1,226,898
TNLTA 12.91 12.92 14.02
TROA 1.08 0.76 0.95
TROE 11.82 7.43 11.58
TEA 9.94 9.00 8.63
TCDTD 0.42 0.00 0.91
TNPAA 0.65 0.06 1.65
TOEAA 2.81 1.40 3.08
TER 63.38 58.00 65.08
TNIIAA 0.73 0.05 1.05
TPPL 146.54 200.05 196.69
TNB 30.15 6.00 112.04

CLUSTER # 4 5 6
FREQ 2 68 86
BUYER
BTA (000) 1,454,248 14,946,683 4,197,501
BNLTA 14.19 16.52 15.25
BROA 0.52 1.13 1.24
BROE 6.72 14.94 14.86
BEA 7.83 7.58 8.38
BCDTD 0.98 0.94 0.92
BNPAA 4.21 1.08 0.71
BOEAA 4.24 3.94 2.87
BER 78.02 68.69 58.69
BNIIAA 1.30 1.86 1.06
BPPL 477.14 151.39 170.12
BNB 47.00 289.10 117.75
TARGET
TTA (000) 198,789 1,088,161 442,413
TNLTA 12.20 13.90 13.00
TROA 0.80 0.93 0.95
TROE 8.88 11.61 11.02
TEA 7.91 8.28 8.95
TCDTD 0.96 0.89 0.89
TNPAA 4.13 1.82 0.89
TOEAA 3.38 3.04 2.88
TER 76.30 67.54 66.13
TNIIAA 0.61 0.80 0.71
TPPL 477.14 151.39 170.12
TNB 6.50 53.30 15.58

FREQ Number of banks in cluster
TA (000) Total assets in thousands of dollars pre-merger
NLTA Natural log of total asset pre-merger
ROA Return on assets
ROE Return on equity
EA Ratio of equity to total assets
CDTD Ratio of core deposits to total deposits
NPAA Ratio of non-performing assets to total assets
OEAA Ratio of operating expense to average assets
ER Efficiency ratio
NIIAA Ratio of non-interest income to average assets
PPL Ratio of price to market price 1 year prior to date
 (target purchase price/lag(tp))
NB Number of branches