 |
| Assets <$10 Million |
I know that some of you are only interested to see if your institution is listed. If you can't wait to know, then find the correct size of your institution at the end of 2010, and click on the appropriate list. I have only listed what I have calculated to be the top 30 in each class. After you see if your institution is there (and therefore whether you agree or disagree with me) please come back and read about how I determined the rankings.
So far, I have discussed how to assess your credit union’s performance in comparison with itself, and in comparison with the industry. But that doesn’t allow the supervisory committee to assess the effectiveness of management in comparison with other credit unions in any given year. To assess performance in credit unions is somewhat difficult, since the only way owners can get a return for their investment is directly, through higher deposit rates or lower loan rates, or indirectly, through improved services for lower fees.
 |
| $10 Million<Assets<$25 Million |
Traditionally, financial institution performance has been assessed using return on assets (ROA). Some early researchers used ROA as a measure of performance, and some still use it as a proxy (in regression analysis only, along with other variables). But ROA does not measure anything directly relevant to the performance of credit unions. It appears to be associated with both high and low performing institutions. The institutions that return high amounts to their members as described above, experience higher demands for deposits and loans which in turn leads to higher demands for retained earnings to meet capital requirements of the regulator. The only way those higher levels of retained earnings can be achieved is through also improving ROA. Those institutions must therefore be extremely efficient in their use of all of their resources.
However, if an institution has poor loan quality, for instance, their loan rates will increase to compensate. If, at the same time, their capital is in danger, they will have to reduce their deposit rates as well. This double hit to members, some of whom were not the cause of their woes, will demonstrate the low performance of the credit union, but if they are successful in earning higher income (to replace capital squandered through poor loan selection or monitoring), the result will appear as high ROA. Both this low performing credit union, and the high flyer described in the earlier paragraph would be listed as high performing.
 |
| $25Mill<Assets<$50Mill |
As academics discovered this discrepancy, they sought ways to better measure performance in credit unions. Actually, some of us had a problem with ROA from a philosophical standpoint. The theory of the credit union suggests that members seek to improve their loan and deposit rates only (only later did researcher acknowledge the importance of services). ROA doesn’t measure anything of most importance to members. Two approaches have been developed to try to address the performance measurement problem.
First, a data envelopment analysis (DEA) approach was developed. This statement is a bit of an understatement. Since all of these methods are not technically DEA, and there are more than one proposal. But they are all closely related in form and function. They all attempt to measure the return to the member by measuring the change in relative efficiency to some standard. They are all “non-parametric” meaning they don’t assume a specific probability distribution. They allow researchers to use as many inputs as seem reasonable, deposit rates, loan rates, services, etc. Efficiency is assessed by looking at how many services can be provided for the total costs to the institution.
 |
| $50Mill<Assets<$100 Mill |
But applied DEA types of procedures have their drawbacks. First, the services tend to be a list, and therefore adding one cheap service is assessed as being equal to adding an expensive service, no matter how much more the member likes the more expensive service. Second, the relative efficiency of any institution is based on the efficiency (as described above) of the most efficient institutions along what appears to be an efficiency expansion path. These most efficient institutions are determined by an effective “dot-to-dot” of efficient institutions, the most extreme institutions that connect are the model institutions. If any of them are themselves extreme, even for a year, they could conceivably affect the “relative efficiency” measure used by the researcher.
 |
| $100 Mill<Assets<$500 Mill |
Because of these problems, and because I wanted to go back to the base theory, I proposed a parametric, and non-parametric approach to detecting abnormal performance improvement (see Bauer, 2008)
[i]. I propose using only the deposit and loan rates – essentially leaving out the services entirely. The services present too many problems in my mind. It is impossible to assess how important they are to members, and therefore how they should wish to substitute one for another. They are certainly not all equal in the eyes of members. The theory does suggest that members should want higher deposit rates, and lower loan rates. With my approach, an expected deposit and loan rate vector is predicted based on how all credit unions adjusted their deposit and loan rates from the last year. The rates last year should be good predictors of this year’s rates because the loan portfolio make-up generally doesn’t change, the hypothesized borrower/depositor preference that may exist among some credit unions can also be preserved, and appropriate market rate adjustments can be made to the rates. Other predictors are added to as control variables. These variables include asset growth during the year (with more assets, the credit union will have to retain more), equity to assets ratio (the lower this ratio, the more the institution will have to retain), and the size of the credit union (measured as the natural log of the ending assets). A tacit assumption in my model is that as performance improves, credit unions tend to spread the gain among all things desired by members – deposit rates, loan rates, services, etc. Therefore, even though services are not specifically measured, they are covered.
For my method to detect higher performance, both the deposit rate has to be higher than would have been expected, and the deposit rate has to be lower than expected. These are measured as abnormal net gain (ANG) – this term and theory were first proposed by Smith, 1980
[ii]. I define abnormal net gain for depositor as
ANGDPY=(DPY – E[DPY]), where DPY is the deposit yield (see last blog entry for calculation particulars). The expected DPY is not the industry average, but an expected rate based on the firm’s own DPY from last year, adjusted for the market adjustments (based on how all other credit unions adjusted their rate), last year’s loan rate (
LIY), the firm’s asset growth, the firm’s capital ratio, and the firm’s size. The other half of the ANG is the abnormal net gain for borrowers,
ANGLIY=(E(LIY) – LIY). No, that is not stated backwards. Members want lower loan rates. They would receive an abnormal net gain if the expected rate were higher than the actual rate. The
LIY is the loan interest yield (see last week’s blog entry for calculation details). Like the expected
DPY, the expected
LIY is not an industry average, but an expected rate based on the firm’s own
DPY from last year, adjusted for market adjustments, last year’s
LIY (based on how all other credit unions adjusted their rate), the firm’s asset growth, the firm’s capital ratio, and the firm’s size.
 |
| $500 Million<Assets<$1 Billion |
Unlike the DEA types discussed earlier, my method produced both a parametric and non-parametric measure. It is the parametric measure that I am employing to generate a ranking of best performers. To qualify as a top performer, both the ANG
DPY and the ANG
LIY must be positive. Because a bivariate regression is used to generate the estimates and expectations, statistical distance from the expected line to the point can be calculated using the chi-square (
c2) distribution.
There is still a drawback to the use of this method to rank credit unions. If an institution just has an outstanding figure for, say ANGLIY, but just a slightly negative figure for ANGDPY, it won’t register as an improvement in performance. This is not a major problem for the use that is usually made of the methodology, since as academic researchers we want to know if a certain event or concept has caused changes in performance, and for every institution with slightly low ANGDPY, there is bound to be another who might have higher than expected ANGDPY, and the methodology would average all institutions with the same event or concept. In a future blog I hope to use this same concept to average the ranks of institutions over years to generate a long-term measure and ranking. But until then, this is my first attempt to apply the system to ranking single year performance improvement.
 |
| Assets>$1 Billion |
In the side of this blog entry are the top 30 credit unions ranked by improved performance from 2009 to 2010. Note, they may have improved because they just did extremely poorly in 2009, or because they are just exceptional. I wanted to, but did not, attempt to pull out any institution whose capital ratios put them in danger of NCUA taking them over. If you don’t see your institution on the list, but would like to know where it fell, please write and ask, I will try to accommodate your request.
[i] Bauer, Keldon, 2008. Detecting abnormal credit union performance.
Journal of Banking and Finance 32, 573-586. I should mention that
Journal of Banking and Finance is a top 10 journal, which is saying something since there are nearly 800 peer reviewed economic/finance journals. That places it in the top 2%.
[ii] Smith, Donald J., 1984. The theoretic framework for the analysis of credit union decision making. Journal of Finance 39, 1155–1168. I should note that the Journal of Finance is the premier academic journal in the field of finance.
Comments
Post a Comment