The past decade has witnessed a disturbing trend regarding commercial bank failures. As can be seen in Table 1, the number of commercial bank failures in the United States has escalated dramatically in recent years (FDIC: Failed Banks, 2013). This fact indicates an increasing pressure on bankers and regulators to understand commercial bank profitability and its components so that the determinants of commercial bank profitability can be properly managed. However, bankers and regulators face tremendous complexity when it comes to determining the proper "mix" of loans and investments held by a commercial bank. Further, there is little in the literature that provides a valid framework to assist bankers in using readily accessible data to understand the factors that determine bank profitability or to analyze those factors for a specific bank. Thus, the purpose of this paper is to present a methodology that commercial banks may use to determine the factors influencing their net profit fluctuations and to demonstrate the use of that methodology for a specific community bank.
Community banks have had a continuous series of challenges and changes during the past decade. Including major regulatory pressures, technological improvements in processing and service delivery, and extreme economic swings in several of the customer segments (DeYoung & Duffy, 2004). The impact of these issues has resulted in Federal Reserve interest rate adjustments, purchase of various asset classes in order to stimulate economic growth and the TARP program to assist in supporting the capital levels in nearly 700 financial institutions according to a Board of Governors of the Federal Reserve System International Finance discussion (2012). The result is a series of internal challenges and a changing paradigm in profitability do to pressures on management and boards, Capital, asset valuation and liquidity (McTaggart & Callaghan, 2011). Community bank CFOs are faced with increasing efforts to support decisions to regulatory bodies, deal with the changing values of asset in loans and investments, and the pressure on actual profitability (Cochero, 2006).
The external environmental dynamics are further driving the need for the study of accurate metrics and understanding of profitability. A group of elite academics met and formulated a number of recommendations for the financial industry at Squam Lake in 2008 (French et al. 2009). Attention to the "too big to fail" segment of the industry drove many of the recommendations but the level of impact on the smaller community banks was for the most part overlooked. The closing of 406 banks since 2008 has included primarily community banks that could not respond to the dramatic turbulence in the industry (Federal Deposit Insurance Corporation, 2013). The opportunities for the smaller bank segment has been focused on responding to the external issues with lesser amounts of attention paid to the possible shifts in internal financial ratios and determinants of profitability (Cocheo, 2011).
Motivation for the study was the pressure experienced by community banks as the result of economic pressures and more specifically mortgage and residential development loan failures. Community banks have well developed internal loan policies and procedures for loan evaluation. The recent loan failures have indicated many banks do not account for the external risk in determining loan approval. Consequently, if appraisal values are increasing quickly collateral of the loan and the financial net worth of the borrowers can be inflated. The increasing values could affect the loan policy calculations. This study considers additional external variables that may affect bank net income.
The community bank selected for this study is a full service $230 million asset institution. The scope of the organization includes five branch locations and approximately 56 fulltime employees. The privately held organization has 108 years of continuous operation with 346 shareholders and no majority ownership. The board membership is of a mix of individuals with agriculture, construction, academic and investment services backgrounds.
Data relating to the internal factors were furnished by Citizens National Bank, Paris, Illinois through income statements and balance sheets, which are prepared monthly, for the period of January 2000 to April 2009. Gross Domestic Product data were obtained from the National Bureau of Economic Research website (www.nber.org). Edgar County unemployment rate data and Consumer Price Index data were obtained from the US Bureau of Labor Statistics website (www.bls.gov/bls/unemployment.htm and www.bls.gov/cpi, respectively). Figure 1 lists the internal and external independent variables that were investigated for explanatory power with the dependent variable, Net Income.
Pair-wise correlations with the dependent variable, Net Income, were calculated to find the specific form of the variable that had the highest possible pair-wise correlation with the dependent variable. Also, lagged values, change values, and lagged changed values of the independent variables were treated the same. The candidate forms of the independent variables were then subjected to further analysis using column correlation vectors in order to mitigate potential multicolinearity problems prior to their inclusion in regression runs. Based upon the column vector analyses, two regression models were rationalized (independent variables were selected for inclusion in the models). The insignificant variables from these runs were discarded and the simplified models were run to obtain improved results. Thus, two models were developed that could be rationally defended, beyond the mechanical approaches.
Examination of Individual Variables: The Search for the Greatest Explanatory Power.
The first task was to identify candidate forms of the independent variables for inclusion in regression modeling. The study began by investigating the pair-wise correlations of each independent variable with the dependent variable, then lags of each independent variable, as well as investigation of correlations of the dependent variable with the change (either increase or decrease) of each independent variable, as well as lags of these changes. Analysis of these pair- wise correlations resulted in a specific form for each of the 16 independent variables that had the highest explanatory power for Net Income. Each of these was considered as a candidate for possible inclusion in further steps of the research method using regression analysis. The following is one example of this best-form identification procedure.
Example of Identification of Best-Form Candidate Variables using Pair-Wise Correlations.
The best form of the independent variable Commercial Term Notes was determined by examination of the pair-wise relationship between the dependent variable Net Income and different forms of the independent variable Commercial Term Notes. The correlation value (r) between Net Income and Commercial Term Notes is .461 ([r.sup.2] is .213). Other correlations between Net Income and changes in the magnitude of Commercial Term Notes (either increases or decreases), including lagged changes that occurred in previous terms, are very small and not meaningful. However, Net Income is reasonably correlated with monthly lags in Commercial Term Notes. The correlation of Net Income with lag one to lag twelve of Commercial Term Notes ranges from a low of .482 at lag one to a high of .569 at lag six, increasing steadily from lag one to the lag six value, then decreasing. Therefore, lag six of Commercial Term Notes was retained. A similar procedure was employed for the remaining 15 independent variables that resulted in identification of the best form of each variable for possible inclusion in regression runs.
Summary of the Best Forms of the Independent Variables.
Table 2 shows the forms of the 16 independent variables that were identified in the above process that were determined to be candidates for inclusion in multiple regression runs. The variables appear in rank order from highest (in absolute magnitude) to lowest correlation values.
A total of sixteen independent variables were examined (three of them are exogenous and the rest endogenous). The values presented above are the correlation values (r) and the coefficient of determination ([r.sup.2]) between the candidate forms of the independent variables and the dependent variable, Net Profit, ranked by explanatory power. The sum of the [r.sup.2] values is much greater...