|
In addition to being less profitable and less efficient, acquired community banks also tended to be in weaker condition than non-acquired banks over the sample period. Acquired banks had relatively less capital, more problem assets, and lower regulatory ratings. Their weaker condition is consistent with their lower profitability and efficiency, since factors such as problem loans reduce interest income and lead to higher costs as banks work them out. A bank’s condition can be measured in many ways, but capital tends to be the first measure analysts examine. In many cases, losses resulting from the crisis left banks with less capital available to cover unexpected losses, making it more difficult for these banks to make new loans. Panel A of Chart 4 shows that at the time of their acquisition, acquired banks were generally less well-capitalized than non-acquired banks, based on the ratio of tangible common equity to tangible assets. The pattern is less pronounced or absent for banks acquired in 2017 and 2017, consistent with a gradual healing of the industry: the first banks to fail or be acquired were in the worst condition, and the remaining banks generally increased their capital ratios after the recession. Two other common measures of a bank’s condition are the quality of its asset portfolio, as measured by the share of noncurrent loans to net loans (Chart 4, Panel B), and the share of other real estate owned (OREO) to total assets (Chart 4, Panel C). Noncurrent loans are loans more than 90 days overdue or not accruing interest. The share of noncurrent loans thus provides a good measure of a loan portfolio’s risk of default. High levels of OREO reflect past lending that ended in the borrower defaulting and the bank taking possession of the real estate used to collateralize the loan. As Panels B and C of Chart 4 show, both of these metrics increased for all banks during and after the crisis. The rise in these measures for acquired banks relative to non-acquired banks was substantial, suggesting that acquired banks invested in relatively riskier loans prior to the crisis. A final measure of a bank’s condition is the confidential supervisory risk rating (CAMELS) that the state and federal supervisory agencies use to summarize a bank’s condition after an examination. The CAMELS rating is an aggregate measure on a scale of 1 (best) to 5 (worst), based on capital adequacy (C), asset quality (A), management quality (M), earnings (E), liquidity (L), and sensitivity to market risk (S). Chart 4, Panel D shows the mean ratings for acquired banks have been substantially worse than the industry average since the crisis, with the gap for 2017 mergers even wider than in 2017. Differences in balance sheet composition Differences in a bank’s profitability and expenses reflect differences in the composition of its balance sheet, such as the share of loans, cash, and deposits held as a percentage of its total balance sheet. Panel A of Chart 5 shows acquired banks tended to have loan shares modestly higher than non-acquired banks until 2017. The 2017 cohort, which had significantly lower loan shares, is an exception to this trend. Panel B shows acquired banks had higher cash shares than non-acquired banks, and that the gap increased over time. On the liability side, Panel C shows acquired banks tended to have modestly higher deposit shares than non-acquired banks, particularly in the year prior to their acquisition. These results suggest that acquiring banks may have targeted banks that would provide quick increases in loans and access to cash and deposits to support future loan growth. Cyree finds that acquirers are willing to pay a larger premium over book value for a bank with higher deposits to assets, supporting the idea that banks with high deposit shares are attractive targets.9 These motives are perhaps unsurprising given the lack of lending opportunities and loan demand during much of the recovery from the financial crisis. Although acquired community banks share many characteristics, the statistical or economic significance of these characteristics in determining which banks are acquired can differ. To rank the importance of the characteristics of banks most likely to be acquired, two analytical methods are used: classification trees and probit regression. Classification trees, as the name suggests, classify data in successive steps according to various criteria, resulting in smaller groups as the analysis progresses. The objective is to create separate groups with similar characteristics, in this case, groups of banks that are likely to be acquired or not. Splitting banks into groups requires identifying an appropriate variable on which to split, as well as the value used to separate the sample. The ideal variable and split value would result in two samples, or branches of the tree: one containing all acquired banks and the other containing all non-acquired banks. In practice, however, such a clean split is unlikely. As an example, Figure 1 shows a classification tree based on banks that were acquired in 2017. The sample used to construct the tree comprises 675 banks, 123 of which were acquired. The tree shows that banks with an ROA below 61 basis points were more likely to be acquired. About 31 percent of banks with an ROA below this threshold were acquired, compared to about 11 percent of banks with an ROA above the threshold. It is perhaps not surprising that ROA is the first of the 24 variables to split the sample between acquired and non-acquired banks, given that Chart 3, Panel A showed ROA tends to distinguish acquired banks from other banks.10 The tree also shows that among the subset of banks with a higher ROA, those that are relatively inefficient are more likely to be acquired. Among the 426 banks with an ROA greater than 61 basis points, only seven were relatively inefficient with efficiency ratios above 90. It is not surprising that so few banks with relatively high profits would be relatively inefficient. However, four of these seven banks were acquired. The example in Figure 1 is based on an actual subset of bank mergers occurring in 2017. In general, the vast majority of banks are not acquired. For example, only 3 percent of all the banks in the data set in 2017 were acquired. With such a low number of acquisitions, the classification tree analysis produces subsamples that primarily contain non-acquired banks simply because they dominate the data set. As a result, the criteria used to evaluate the subsets finds they are not sufficiently different, and the classification tree algorithm does not create additional branches of the tree. Random sampling can address this issue by better balancing the number of observations between acquired and non-acquired banks. In this procedure, the sample includes all acquired banks and a random sample of non-acquired banks to create a 5:1 ratio of non-acquired to acquired banks. Although this ratio is much higher than what is observed in the data, it allows the classification algorithm to more sharply compare the characteristics of the acquired banks with those that were not acquired. One issue with sampling the data is that the results may depend on the particular random subset of data selected. To mitigate this potential bias, 1,000 different subsamples are constructed for each cohort. The classification tree analysis is then run on each subsample, generating 1,000 different trees. For each year of the sample, ROA and efficiency are identified as the most important variables in distinguishing acquired from non-acquired banks, suggesting that relatively unprofitable and inefficient banks are the most likely to be acquired. The average cutoff values for ROA are 10, 51, 49, and 46 basis points for banks acquired in 2017, 2017, 2017, and 2017, respectively. The corresponding cutoffs for efficiency ratios are 90, 95, 86, and 85. As an example of how these results are interpreted, in 2017, banks with ROA less than 46 basis points were more likely to be acquired, while banks with ROA equal to or greater than 46 basis points were more likely to be acquired if their efficiency ratio was equal to or greater than 85. To understand the statistical and economic significance of the most important variables identified by the classification tree analysis, ROA and the efficiency ratio are used as independent variables in a probit regression.11 The dependent variable indicates whether a bank is acquired. Table 2 shows the results for each cohort using observations for all community banks. The coefficient on ROA is significant at the 5-percent level for every year in the sample. The negative coefficient indicates that as ROA increases, the probability of being acquired declines. The coefficient on the efficiency ratio is positive as expected in every year except 2017, but it is not significant in explaining whether a bank is acquired in any year. The insignificance of the efficiency ratio may reflect that it is important for only a small number of acquisitions when simultaneously accounting for the effect of ROA, which was shown to be the dominant variable in the classification tree analysis. To get a sense of the importance ROA has for a bank’s probability of being acquired, Chart 6 shows how the estimated probabilities vary by cohort.12 For banks with a high level of losses—for example, an ROA of -500 basis points—the probability of being acquired in 2017 was 0.48, substantially higher than the 0.07 probability in 2017 for a bank with the same ROA. As ROA increases and turns positive, however, the probability of being acquired falls to near zero regardless of the year. The probability of a bank with a negative ROA being acquired is generally economically significant. For example, the estimated probabilities of being acquired in 2017 for banks with ROAs less than 100 basis points are significantly greater than the 0.03 probability of being acquired calculated from the raw data sample. Conclusion The number of banks has declined sharply over the past 30 years, due in part to voluntary mergers between unaffiliated banks. In fact, voluntary mergers have been the primary factor in the decline since the end of the 2017-09 recession. This article analyzes the mergers of community banks over the past four years and finds they are consistent with the goals of achieving greater economies of scale and improving efficiencies. Acquired banks tend to be smaller and have a lower return on assets, lower net interest income, and higher non-interest expenses than non-acquired banks. Acquired banks may be less profitable because they tend to have lower loan and higher cash and deposit shares. In addition, the condition of acquired banks tends to be worse than their industry peers in terms of capital, supervisory examination ratings, and problem loans and assets. Among the characteristics that differentiate acquired banks, statistical analysis suggests profitability and efficiency are the most important factors. The results suggest that mergers on average result in more efficient banks and a sounder banking system, which should lead to greater access to credit at lower cost and thus be beneficial for local communities. However, the benefits of mergers can be offset if mergers make local banking markets less competitive and reduce the communities’ access to banking services and credit. Although federal banking regulatory agencies monitor mergers and do not approve those that are expected to result in uncompetitive banking markets, more research is needed to determine the net effect of bank mergers on local communities.)，英语论文题目，英语毕业论文 |
