Is the Size of the Loan Portfolio Important to Profitability?

Published in Topics: Bank Analytics & Performance, Board & Executive Education

• Print
• Send to a friend
• Comment
• Bookmark
• Recommend
   (6)

Executive Summary

Recent research conducted by BNKAnalytics led to two important conclusions:¹

• Banks with larger loan portfolios tend to produce higher returns on average equity (ROAEs) than banks with smaller portfolios.
• However, when we controlled for the size of the loan portfolio—in terms of total loans—the portion of loans didn’t correlate to performance.

How does a bank manager use these results? These suggest that as a bank grows in size, the bank can look forward to incremental improvements to ROAE. The first conclusion points to a synergistic effect for banks. Many bankers will see this conclusion as a confirmation of what they already believe.

The second conclusion is more interesting. It suggests that a bank should not concern itself with the portion of loans versus investments on the balance sheet. The average portion of loans for banks under $3 billion is 64 percent of assets. If a bank has a much higher portion of loans, that doesn’t mean it will be more profitable, and vice versa. Some bank managers and board members will find this surprising.

The Strategic Implication of our Research

During a strategic planning session, BNK was working with a bank’s board of directors, looking for ways to reverse an alarming downward trend in the bank’s performance. In past years, the bank’s ROAE was consistently above 15 percent. The board was concerned and puzzled because performance was slipping, despite a high loan-to-assets ratio (L/A). They felt that the bank’s large portion of loans would ensure continued success. Yet, that wasn’t the case.

Obviously, there are pricing- and credit-related factors that can hamper the performance of a bank’s loan portfolio, so no banker would be surprised by the fact that a large loan portfolio doesn’t necessarily lead to high profitability. But, it does raise an interesting question. Generally, are banks with larger loan portfolios more profitable? We often hear bankers talk about their intention to limit the size of the investment portfolio, as though a larger portion of investments will hurt performance. Bankers who strive for a large loan portfolio or a small investment portfolio are actually pursuing the same goal—they are just looking at it from a different angle. Should performance-driven banks pursue a target for the size of their portfolio?

Another bank comes to mind in the context of this research. During a strategic planning ses-sion, the CEO of a mutual savings bank observed that his bank’s investments represented more than 50 percent of its assets. So, the loan portfolio is smaller than the investment portfolio. Despite the bank’s mutual status, the CEO’s goal is to move the bank closer to the BNK Elite’s level of perform-ance, which is typically an ROAE around 15 percent. He wondered whether it would be necessary to shift a portion of the bank’s investments into loans in order to meet his higher ROAE goal.

Our research suggests that neither bank described above should focus on the portion of loans as it addresses its performance concerns. If a bank can grow through an increase in the size of its loan portfolio, that growth can produce synergies and improve the bank’s ROAE. On the other hand, to strive for a higher portion of loans as a strategic goal is unnecessary and, perhaps, ill-advised.

Description of Analysis

The remainder of this paper is for those curious as to the research, statistics and analysis be-hind our conclusions. We have included a discussion on the research methodology used. The data used for the study was obtained from SNL Financial. The econometrics and interpretation was done by BFRC Services, a firm that specializes in economics and financial research.

For years, the loan-to-deposit ratio (L/D) was the measure used by banks to compare the relative size of their loan portfolios to other banks. Today, borrowings from the Federal Home Loan Bank (FHLB) are often a significant portion of funding for banks. Therefore, it is common to see a bank with an average-sized loan portfolio have a high L/D ratio. Thus, when examining the size of the loan portfolio across banks, it is arguably more appropriate to look at the L/A ratio.

The performance of a bank depends on so many factors, including:

• the yield on earnings assets, such as loans and investments
• the cost of funds, including borrowing costs
• net overhead, the difference between noninterest income and noninterest expense
• the tax burden
• the balance between earning and nonearning assets and asset allocation; and
• the amount of financial leverage.

It is impossible to hold all factors constant and vary the size of the loan portfolio in isolation. Therefore, in order to uncover any relationship between the size of the loan portfolio and bank performance, econometric techniques must be employed.

What Was Tested?

We use bank data provided by SNL Financial spanning 1990 through 2003 for total assets, L/A ratios, and ROAEs. BNK’s niche is the community bank sector, so we look at banks that are $3 billion and smaller. Our first objective is to determine whether or not there is a direct relationship between the size of the loan portfolio and performance, as measured by ROAE. The second objective is to uncover whether a relationship exists between the L/A ratio and performance.

To test the second objective, we hold constant the size of the loan portfolio. In other words, to test the link between the L/A ratio and performance, our procedure is to test similarly sized banks to other similarly sized banks.

Econometric Methodology

A straightforward methodology commonly employed to control the influence of a variable in financial research is to simply break the data up into roughly equal “buckets.” The idea is to be able to compare certain characteristics while holding a variable relatively constant. Think in terms of NFL football players. Would it be fair to compare running backs with offensive linemen in the 40-yard dash? Probably not. So, we would want to break players out by position (the influencing vari-able in this case) and then compare 40-yard dash times.

The first step is to separate the data into five equally sized buckets based on loan size. The Summary Table is shown in Figure 1. In technical terminology, these “buckets” are called quintiles (i.e., five segments) and illustrated vertically in Figure 1. Each bucket contains between 3,400 and 3,900 observations with relatively similar loan size portfolios.²

Figure 1: Summary Table

Bucket 1 (i.e., quintile 1), the upper most row in Figure 1, consists of data from banks with the smallest loan portfolios, and bucket 5 consists of data from the banks with the largest loan portfolios. The average loan portfolio size increases sequentially as we progress from the top row down to the bottom row. In other words, these loan bucket sizes are listed vertically in ascending order in the Summary Table. By breaking the data up in this way, we can better control for loan size on other variables.

Now, we take each loan size bucket and break it down into five equally sized buckets based on loan-to-asset ratios. The first bucket consists of data from banks with the smallest L/A ratios, while bucket 5 contains data from banks with the largest L/A ratios. The L/A buckets are listed horizontally at the top of Figure 1. As we move from left to right, the average L/A ratio of each resulting bucket increases.

This process produces a five-by-five checkerboard that contains a total of 25 buckets, or cells, as illustrated in Figure 1. Finally, we take each of these checkerboard cells and compute summary statistics on all the banks in each cell—specifically, we compute the average ROAE and the standard deviation of the ROAEs.³ If loan size or L/A have no effect on performance, we would not ex-pect the average ROAEs to be any different as we move from cell to cell in our checkerboard.

Clearly, as we move up and down in each L/A column, the average ROAEs are not equal. In fact, without exception, as loan size increases, so does performance. Conversely, as we move from left to right in each row, the average ROAE does not change meaningfully. These qualitative findings suggest that loan size certainly affects performance; more specifically, our findings indicate that the larger the loan portfolio, the higher the ROAE. Controlling for loan size, however, the size of the L/A ratio does not appear to have a meaningful impact on performance.

Are the Results Robust?

To be complete, we try to quantify how “good” or “meaningful” our findings are within a statistical context. That is, how confident are we in stating that loan size does indeed affect ROAE? Using what we call a “t-statistic” we can evaluate whether the difference in average ROAEs are “statistically different” from cell to cell. The larger the statistic, the more confident we are that the ROAEs are statistically different from one another.

Figure 2 provides the t-statistics comparing the ROAE for each bucket to the others by loan size. Generally, a t-statistic higher than two reflects significance. The t-statistics between cells of different loan sizes (i.e., cells from different rows in the table) imply incredibly high statistical significance. Indeed, all are significant at the 99-percent confidence level.

Figure 2: T-Statistics for Loan Sizes

* All are significant at a 99% confidence interval

Similar to the format used in Figure 2, the data in Figure 3 are the t-statistics comparing the ROAE for each bucket to the others, by the L/A ratio. In contrast to the last comparison, the t-statistics between cells of different L/A ratios (i.e., cells from different columns in the figure) rarely imply any statistical significance. Only one L/A comparison is significant (i.e., bucket 3 versus bucket 5). Results are consistent with our original observations from the Summary Table.

Figure 3: T-Statistics for Loan/Asset (L/A) Ratios

* Significant at a 99% confidence interval

In summary, we have shown that performance is significantly different depending on the loan portfolio size. Specifically, as the loan size in total dollars increases, so does performance. Moreover, the results are statistically robust. However, controlling for loan size, performance does not seem to be affected by the portion of loans on the balance sheet, as measured by the loan-to-asset ratio.

¹ Walker’s coauthor, Gerald “Jeff” Buetow, is the president of BFRC. Jeff is the founder of BFRC Services, LLC, a Virginia-based financial research consulting firm. He previously served as vice president of curriculum development at the CFA Institute. He also served as the Wheat first professor of finance and director of the Quantitative Finance program at James Madison University.

² The reason that the number of observations (N) is not exactly equal in each bucket is due to missing ROAE values. For example, if a bank contained loan size information but no ROAE information, it would not show up as an observation within a bucket, but would have influenced the creation of the loan bucket. Since missing values are ignored in the computations, they do not affect the analysis in any way.

³ The average ROAEs are in the columns headed with the word “mean.”