Exploring Banking Performance: Where are the Limits?

Abstract

  

The relationships between non-performing loan percentage, net interest margin, leverage and the ROE and ROA of 284 banks are explored. Using ten-year averages (1983 - 1992), the nonlinear complex performance relationships are explored and specified, resulting in higher explanatory power than using linear estimations. Resulting R^sup 2^'s are .497 for ROE and .644 for ROA. For these data, during this time, in order to maximize performance, non-performing loans should be reduced to the lowest limit and net interest margin should be targeted at one standard deviation higher than the mean of all banks. 

Introduction 

Since the early 1990's the banking environment in the US has undergone a number of dramatic changes. Extensive consolidation has taken place. That and the explosive growth in computing power available to users has led to significant automation of a number of banking practices. (Carter and McNulty, 2004) One example of a process that has become more data driven and automated is the use of small business credit scoring (SBCS) for loans to small businesses (Berger, Frame, and Miller, 2005).

In their study of banking strategy and its implementation Hatten, James, and Meyer (2004) found that the implementation was a far more powerful determinant of performance than strategic choice of competitive posture. They did not explore the exact specifications of the relationships between measures of implementation and the resulting performance. That is the focus of this paper, which examines the loan process to better specify the impact of its implementation on bank performance.

The article proceeds as follows. Section 1 provides the background of the banking environment wherein the decisions regarding the loan process and leverage take place. Section 2 provides a development of the related literature resulting in presentation of the hypotheses. Section 3 discusses the data sources. Section 4 describes the variables, gives sample statistics, and presents the empirical models. Section 5 presents the empirical results. Section 6 concludes.

Background

The banking literature in the early 1990's focused on the empirical testing of the effects of bank concentration and competition using the traditional structure-conduct-performance (SCP) hypothesis on U.S. banking industry data. The SCP hypothesis claimed that bank concentration and other impediments to competition create an unfavorable environment for bank operations from a social viewpoint. This research usually specified bank prices and measures of profitability as the endogenous measures of bank conduct and performance. The focus was on the banking industry as a whole, not individual banks and their performance.

During the same time, the business strategy literature, drawing on Porter (1986), sought empirical evidence to specify sources of competitive advantage (e.g., McKee, Varadarajan, and Price, 1989; Delery and Doty, 1996). This research sought to find factors leading to advantage that would generalize across industries. The thrust was to determine the factors, not to specify, precisely, their relationship to performance.

Berger, Demirguc-Kunt, Levine, and Haubrich (2004), point out that the banking literature has now moved to incorporate aspects from the strategy literature. More recent research has moved beyond the SCP hypothesis, and examined a number of different models of competition. The measures of conduct and performance that are analyzed have grown to include various indicators of efficiency, service quality, and risk of the banks, along with consequences for the whole economy.

The lending process is a key component of banking performance. Recent work (Berger, Frame, and Miller, 2005; Carter and McNulty, 2004) has focused on the changes that have occurred in this practice over the last decade. An assumption, supported by data, in these studies is that differences in the lending process across banks lead to performance differences. The present study shows that these differences existed prior to the changes in the banking industry and provides specific targets for achieving better performance.

Recent Literature, Development, and Hypothesis

Within the heavily constrained banking environment, there are limits, both upper and lower, to achievable performance. For instance, higher interest rates may be offered to depositors to attract those deposits, but doing so may decrease return on equity, making a bank choosing and implementing that strategy less attractive to investors. (This does not apply to closely held banks that do not expose themselves to capital market pressure.) Thus, for a given aspect of implementation, say interest offered, banks vary from the minimum to the maximum within a tight range.

As banks make decisions and implement their strategy within this tight range, decision-makers at each bank will establish parameters for decisions in a number of areas. This study examines the decisions resulting in Net Interest Margin, Non-Performing Loan percentage, and Leverage across banks to determine the effects of these variables on performance - measured by ROA and ROE. 

Starting with the interest charged component of Net Interest Margin (NIM) and the Non-Performing Loan percentage (NPL), banks analyze loan applicants as to the level of risk that the loan will default and will make loans within a set of chosen parameters. Those parameters establish the interest rate that applicants will pay for the loan compared to the risk calculated by the loan officer - the classic risk/return ratio. As banks use the same variables to evaluate risk level, the only variance will be from the variety of loans offered. If the risk/return ratios are the same across all classes of loans (high to low risk; personal, auto, home, commercial, etc.), across all banks and markets, there will be no opportunity to improve performance over competitors via the loan process and portfolio management.  

Recently, as banks have been increasingly using small business credit scoring (SBCS) for loans to small businesses (Berger, Frame, and Miller, 2005), models that use a standard set of variables, weighted similarly across banks, risk analysis for these loans has become standardized across banks [Footnote on providers and use, see p. 193]. For loan applications analyzed using SBCS, the greater the risk, the higher the interest rate that can be charged. As will be shown, there is no reason to believe that the risk/return ratio is constant - implying a normal return distribution. Further, for loans made across all banks, it is unlikely that the risk distribution of those loans is normal. Examining all loan applications should reveal that the risk distribution is skewed to the high side. Lastly, it is also quite probable that there was more variance in the risk/return ratio across banks prior to the increased use of SBCS due to the higher variance across banks in the set of factors used to determine risk.

Examining the distribution of returns, for any particular loan application and its associated calculated risk, the limits are easily determined. The upper limit to the return is the willingness of the applicant to pay the interest rate - the more interest to be paid, the higher the applicant's income has to be to pay those charges. The lower limit to return is how low an interest rate any bank will be willing to charge for a completely 'safe' loan.

If, in using SBCS, the expertise in calculating risk is equal across all banks - the risk/return ratio will be constant - there should be no differential gain available to any bank from offering higher or lower interest loans - the losses from non-performing loans will exactly offset the gains from the performing higher interest loans. If the risk/return ratio is not the same across loans, that is the interest charged for more risky loans is higher than the average risk/return ratio, banks choosing to make these loans have a source of competitive advantage with respect to ROE and ROA (Berger, Frame, and Miller, 2005). Therefore, putting that set of excess returns in historical perspective is important in determining the extent of risk/return variance that was reduced as SBCS came into wider practice. 

Any bank can strategically choose to make loans of higher risk, charging higher interest rates for that class of loans. Just as any bank can strategically choose to pursue only extremely "safe" loans. Within the banking population, however, banks choosing the latter will suffer in their performance due to the missed opportunity costs (and higher returns) of fewer "safe" loans - at least to the extent that those loans are repaid at a higher than expected rate. Similarly, those banks "pushing the envelope" on loans will have lower performance unless their choice of interest rates for higher risk loans leads to higher average risk/return ratios. Obtaining higher risk/return ratios may be due to the lower competition among lenders for higher risk loans. Thus the expected relationship between the average interest rate charged in a bank's loan portfolio and a bank's performance is expected to be in the shape of an inverted "V" - banks with lower average interest are expected to have lower performance, as are banks with much higher than average interest, at least for a given non-performing loan percentage and a given cost of funds (ceteris paribus).