A survey done a few years ago of the interest rate risk managers for the largest U.S. and Canadian banks confirmed some common elements in their approaches to asset and liability management. First, more than 90% of the respondents used just one yield curve as the “base yield curve” for valuation and profitability analysis. Second, this curve almost universally the U.S. dollar interest rate swap curve. Since that survey was taken, interest rate swap spreads have been negative at many maturities and massive fines have been paid by major banks for manipulation of Libor, the short term interest rate at the heart of interest rate swap spread setting. This note show how to improve the accuracy of transfer pricing and valuation in asset and liability management by eliminating the use of the interest rate swap curve, replacing it instead with the U.S. Dollar Cost of Funds IndexTM. We use traded bond price data for KeyBank N.A. and its parent KeyCorp to illustrate the process.
Conclusion: The interest rate swap curve, long used by many banks as a transfer pricing and valuation yield curve, is no longer credible to bank management, shareholders or regulators. The swap curve has been manipulated on both the short end and the long end, and its negative spreads to Treasuries have been persistent indicators that it is no longer relevant. The U.S. Dollar Cost of Funds Index is a much better basis for transfer pricing. The Index is based on real trades, unlike Libor, and it is directly designed to measure the marginal cost of funds for the largest U.S. deposit-taking institutions.
Background on Traditional Transfer Pricing and Asset and Liability Management Practices
Transfer pricing is the discipline in interest rate management of assigning a “matched maturity cost of funds” to every asset and liability of a complex financial institution. This marginal cost of funds, for internal profitability calculations, represents the cost of borrowing money to fund an asset from the bank’s Treasury or (best practice) transfer pricing center. From the perspective of liabilities, the transfer pricing center “buys” the liability from the originating office. The transfer pricing system was first developed by a team led by Wm. Mack Terry in 1973 in the Financial Analysis and Planning Group at Bank of America Corporation (BAC) in San Francisco. For a history of transfer pricing, see van Deventer, Imai and Mesler (2013), Chapter 2, or this more detailed version of the history of transfer pricing.
The early practitioners of transfer pricing always viewed the transfer pricing yield curve as the best available estimate of the institution’s marginal cost of wholesale funding in large size. In most large banks, corporate loan transactions are priced in real time based on the bank’s cost of funds at that instant. Even 30 years ago, First Interstate Bancorp (then the seventh largest bank holding company in the U.S.) used real time pricing for all intra-company transfers of funds for amounts of $10 million or more. Over time, common practice for all but the largest institutions focused on the interest rate swap curve as a proxy for the marginal cost of funds. A proxy was needed because most institutions do not have actively traded bonds or credit default swaps on their name. A recent study by Kamakura Corporation confirmed that only 14 U.S. bank legal entities have had actual credit default swap trades on their names since 2014, compared with a total of 6,730 banks insured by the U.S. Federal Deposit Insurance Corporation. It was widely recognized that the swap curve was at best a crude approximation of the marginal cost of funds, because interest rate swaps do not involve the exchange of principal and the risk of loss on that principal amount.
Another simplification that became common practice over time was the use of only one valuation yield curve for interest rate risk analysis and mark-to-market calculations. The bank’s transfer pricing yield curve, usually the swap curve, was the most common choice. Bankers recognized that this too was an approximation to the best practice technique, using the IBM yield curve to value loans to IBM, using the primary mortgage market yield curve to value mortgages and so on. This data is now widely available from firms like Kamakura Corporation but the use of a single yield curve is still common practice.
Both of these elements of common practice are now in need of drastic improvement because of changes in the market place since the credit crisis of 2006-2010.
Problems with the Swap Curve
Besides the problems summarized earlier, the interest rate swap curve is simply a poor estimate of the true marginal cost of bank funding, as shown in this data from the Federal Reserve H15 statistical release for August 22, 2014:
Note that the 30 year interest rate swap spread has a margin of only 0.01% over the U.S. Treasury 30 year yield. By contrast, the Citigroup 8.125% bonds due July 15, 2039 had a matched-maturity credit spread over Treasuries of 1.605% and the JPMorgan Chase 6.40% bonds due May 15, 2038 had a credit spread of 1.506% over Treasuries. The interest rate swap spread of 0.01% was off by about 150 times the true marginal cost of funds credit spread. See the appendix for more points of comparison.
The U.S. Dollar Cost of Funds IndexTM
The U.S. Dollar Cost of Funds Index was developed by Kamakura Corporation to specifically address the weaknesses of the swap curve as a transfer pricing or valuation yield curve:
- It does not represent a marginal cost of funding
- The short term interest rate side of interest rate swaps has been manipulated and billions of dollars of fines have been paid
- There are lawsuits and investigations into allegations that the fixed rate side of the swap curve has also been manipulated
- In part because of these reasons, the swap spread to U.S. Treasuries has been negative at many maturities for an extended period of time, further calling the credibility of the curve as a proxy for funding costs into question.
The U.S. Dollar Cost of Funds Index is a composite marginal cost of funds for the 4 largest U.S. deposit taking banks, Bank of America (BAC), JPMorgan Chase (JPM), Citigroup (C), and Wells Fargo & Co. (WFC). Kamakura Corporation uses regression analysis to fit a composite spread curve to secondary bond market trades for these four firms. The U.S. Dollar Cost of Funds Index for August 22, 2014 is shown here:
On August 22, 2014, the big 4 banks whose bond trades underlie the index were all ranked among the 20 most heavily traded issuers in the U.S. bond market, with Bank of America ranked first:
The calculation of the index is explained in more detail in a recent post from Kamakura Corporation. The appendix shows the bond issues used in the calculation on August 22, 2014 for all 4 banks and their bond credit spread differentials versus the composite U.S. Dollar Cost of Funds Index.
We now use the secondary market bond data for KeyBank N.A. and KeyCorp (KEY) to derive a marginal cost of funds index for both the bank and the parent company.
Calculating a Transfer Pricing Yield Curve Using the U.S. Dollar Cost of Funds Index
Most of the institutions who have used the interest rate swap curve as the basis for transfer pricing have assumed that the curve itself is a good enough proxy to the cost of funds that one needs not add or subtract a spread. When using secondary market bond trade information from TRACE and the U.S. Dollar Cost of Funds Index, this common assumption of “no spread” is no longer necessary. On August 22, 2014, KeyCorp and its bank subsidiary KeyBank National Association had some secondary bond market trading activity:
KeyBank N.A. and KeyCorp had a total of $11.56 million in principal amount traded in 11 trades on 4 different bond issues. We can calculate the spread between the traded levels on these bond issues and the matched maturity spreads on the U.S. Dollar Cost of Funds Index as a proxy for the incremental costs of funds for KeyBank N.A. and Key Corp. Those spreads are shown in this table:
After eliminating callable issues and floating rate notes for all financial institutions, we have 1 KeyCorp bond issue and 2 KeyBank N.A. bond issues to work with. This is too little information to build a credible stand-alone yield curve for KeyCorp like we have in this graph for JPMorgan Chase & Co.:
Instead we focus on the gaps between the credit spreads for the KeyCorp entities and the U.S. Dollar Cost of Funds Index. The KeyCorp 3.75% bonds due August 13, 2015 have slightly less than one year to maturity. They have a credit spread of 0.476% on August 22, 2014 versus the matched-maturity U.S. Treasury curve. The matched maturity U.S. Dollar Cost of Funds Index credit spread for bonds maturing August 13, 2015 is 0.466%, so KeyCorp’s bond is trading at just 0.01% above the U.S. Dollar Cost of Funds Index at this maturity. KeyBank N.A. has a bond with a coupon of 5.091% due on March 26, 2015 with a credit spread of 0.347%. This credit spread compares with the U.S. Dollar Cost of Funds Index for bonds due March 26, 2015 of 0.418%, so the KeyBank issue is trading at 0.071% below the U.S. Dollar Cost of Funds Index. A second traded KeyBank bond issue is the 1.65% bonds due February 1, 2018. These bonds traded at a credit spread of 0.501% to U.S. Treasuries on August 22, 2014. The matched maturity U.S Dollar Cost of Funds Index credit spread was 0.740% so this KeyBank issue is trading at a full 0.239% spread below the U.S. Dollar Cost of Funds Index.
How do we use these spreads to the U.S. Dollar Cost of Funds Index? This is where judgment and more data come into play. Based on just one day’s data, as a first approximation, we’d say that KeyCorp trades at a 0.01% premium to the U.S. Dollar Cost of Funds Index. A similar first approximation says the following about KeyBank N.A.:
- The bank is trading at tighter credit spreads to the U.S. Dollar Cost of Funds Index than the parent company.
- For maturities of March 26, 2015 or less, the best available estimate of the spread to the U.S. Dollar Cost of Funds Index is 0.071%.
- For maturities of February 1, 2018 or later, the best available estimate of the spread to the U.S. Dollar Cost of Funds Index is 0.239%.
- For maturities between the two dates, a linear interpolation between 0.071% and 0.239% would be used.
As more data becomes available, these spreads to the U.S. Dollar Cost of Funds Index would be refined and one could decide if using spreads to the index on a lagged basis adds value or not. If a very large block of trade data becomes available (say because of a major news event that triggers trading activity), a stand-alone calculation of the marginal cost of funds curve relative to U.S. Treasuries for both KeyBank N.A. and KeyCorp may be feasible, as it is in the JPMorgan Chase example above.
The interest rate swap curve, long used by many banks as a transfer pricing and valuation yield curve, is no longer credible to bank management, shareholders or regulators. The swap curve has been manipulated on both the short end and the long end, and its negative spreads to Treasuries have been persistent indicators that it no longer relevant. The U.S. Dollar Cost of Funds Index is a much better basis for transfer pricing. The Index is based on real trades, unlike Libor, and it is directly designed to measure the marginal cost of funds for the largest U.S. deposit-taking institutions. We close with a table of the bond issues and their contribution to the U.S. Dollar Cost of Funds Index on August 22, 2014. More information is available from firstname.lastname@example.org.
Date Used in Constructing the U.S. Dollar Cost of Funds Index,
August 22, 2014
Copyright ©2014 Donald van Deventer