The MSR remedy

Hedging of mortgage servicing rights has been a perennial bugbear for risk managers. New US accounting rules have attempted to simplify matters, but some complexity and confusion remain. By Navroz Patel

p38-jpg

As this year's first annual reports begin to be filed later this month, an important new chapter for many US mortgage banks begins. Mortgage servicing rights (MSRs), which are generated when a firm sells mortgages and retains servicing of the loans or enters into a contract to service the mortgages of a third party, have always caused a tremendous headache for risk managers. But last year, the US Financial Accounting Standards Board (FASB) offered a palliative in the form of Financial Accounting Standard (FAS) 156, which covers accounting for servicing of financial assets.

Prior to the release of FAS 156 in March 2006, US mortgage banks were required to mark derivatives to market, while MSRs were accounted for on a lower-of-cost-or-market basis. This created a problematic accounting asymmetry for hedgers. On an economic basis, any losses on MSR hedges would be offset by an appreciation in the value of the MSR portfolio. On an accounting basis, however, the appreciation in the value of the MSRs would be limited to the amortised book value, preventing the full offset of any decline in the mark-to-market value of the hedge.

FAS 156 amends the previous rule on accounting for transfers and servicing of financial assets (FAS 140) by requiring MSR assets and liabilities to be initially measured at fair value. The servicer can then choose either a traditional amortisation method, similar to FAS 140, or a new fair-value method - where changes in fair value are reported in income statements. Furthermore, servicers are free to elect amortisation for some of their assets and the fair-value approach for others.

Commenting at the time FAS 156 was released, the FASB said the amendment to FAS 140 sought to allow for hedge accounting-like treatment without having to deal with the complexities of FAS 133, the standard for financial reporting of derivatives and hedging transactions.

True hedge accounting allows the gain or loss of a derivative to be offset in the earnings statement by the loss or gain on the hedged item. To qualify for hedge accounting under FAS 133, the hedge must be deemed highly effective, both at the inception of the hedge and on an ongoing basis - in other words, the company must prove that the gains or losses on the derivative will largely offset changes in the fair value of the hedged item.

Prior to FAS 156, hedge accounting under FAS 133 had, in theory, been the ideal choice for those firms troubled by the earnings volatility that was sometimes a consequence of the accounting asymmetry under FAS 140. In reality, however, mortgage servicers found it hard to qualify for hedge accounting. Proving the effectiveness of a hedge for a portfolio of MSR assets was difficult, in part because of unpredictable prepayment behaviour of borrowers, while the costs of building the infrastructure to track MSR and derivative values and assess hedge effectiveness on an ongoing basis were high. For these players, FAS 156 offers an effective middle ground.

"FAS 156 is extremely important for the majority of US institutions with MSR portfolios, as it effectively removes the need for hedge accounting," says Kent Westerbeck, Chicago-based group senior vice-president at ABN Amro North America. "Instead, you can put a hedge on to take away the majority of the risk and mark-to-market the difference in value of the hedge and the MSR in the income statement."

Simply put, FAS 156 should allow for the true economics of the relationship between the MSR and its hedge to emerge in the financial statement, says Tim Kviz, a professional accounting fellow in the Office of the Chief Accountant at the US Securities and Exchange Commission (SEC). "It was an extremely cumbersome process to get hedge accounting previously, and some even viewed it as an impossible task," he says.

By adopting FAS 156, companies can implement something similar to hedge accounting, without being required to demonstrate hedge effectiveness as prescribed by FAS 133. Proving the effectiveness of a hedge under FAS 133 is difficult because prepayment behaviour of borrowers means the fair value of MSRs does not move linearly with interest rates - MSRs lose value more quickly when interest rates decline than they gain value when interest rates rise.

The difficulty of achieving hedge accounting for MSRs is also in part tied to a lack of uniformity in the loans that underlie them: loans will perform markedly differently in different interest rate scenarios. Even if the loans are fairly plain vanilla - 30-year fixed-rate mortgages, for example - the propensity to prepay can differ significantly among borrowers.

"There is no good single derivative to hedge MSRs, even when you just consider prepayment propensities in terms of the coupon dispersion," says the SEC's Kviz. "And then, if you consider the fact that changes in MSR value lags interest rate moves, and other non-interest rate-related factors in prepayment speed such as location, the inherent complexity becomes obvious."

To get hedge accounting, servicers would typically split up portfolios into hundreds of buckets of similar assets. They would then assess which buckets should remain unhedged and which should be hedged using swaps, swaptions, mortgage-related securities or some combination of the three.

"Many don't hedge because they think it terribly more complicated than it actually is," says Westerbeck. "It's fairly straightforward to take off 80% of your risk with a hedge. But getting that to satisfy accounting criteria was far from easy before FAS 156."

Even with the introduction of FAS 156, a number of firms have decided during the past 12 months to reduce their MSR portfolios or leave the business entirely, deterred by significant swings in their profit and loss statements in the past (Risk July 2005, page 42-44). Washington Mutual, one of the largest US mortgage banks, with assets of around $350 billion, identified a reduction of its exposure to MSRs as one of its key strategic objectives for stabilising earnings last year. During the third quarter of 2006, it sold around $2.5 billion worth of MSR exposures - nearly one-third of its overall MSR portfolio.

Meanwhile, Atlanta-based NetBank sold 70% of its MSRs in two deals for around $119 million in September 2006. Speaking at the time, the firm's chief executive, Steven Herbert, said the transactions eliminated significant earnings volatility. "We will no longer have the same exposure to impairment and hedge-related losses," he added.

Most recently, Citigroup announced on January 22 that it would buy ABN Amro Mortgage Group, which has a $224 billion mortgage servicing portfolio. The Michigan-based mortgage firm is a subsidiary of Chicago-based LaSalle Bank, itself a US banking subsidiary of Netherlands-based ABN Amro.

Westerbeck declined to comment specifically on whether the accounting regime contributed to ABN Amro's decision to sell its MSR assets. As an international financial institution, ABN Amro also reports under International Accounting Standard 39, which covers the measurement and recognition of financial instruments. As such, ABN Amro typically seeks hedge accounting in the US.

"Accounting is definitely one of the characteristics of MSR assets that is troublesome to a bank's management," he says. "You can spend money to get hedge accounting, but there's always a not insignificant possibility that you fail for a reason today that wouldn't have been a valid reason to fail the week before." (See box: Q&A.)

While most banks have welcomed the introduction of FAS 156, some analysts have warned that giving firms the option of using the fair-value method will make it more difficult to compare financial statements. "We clearly recognise that there is a comparability issue here," says Joyce Joseph-Bell, a New York-based director in the financial services ratings group at Standard & Poor's. "Despite its obvious benefits, introducing the MSR fair-value treatment as an option further compounds the variability found in comparing peer financial reports."

As a result, servicers will need to be more forthcoming in their disclosures in order to give investors and analysts all the information they require, says Joseph-Bell. "We will be looking for detailed information regarding why a company made the choice it did, and the risks inherent in a portfolio," she says. "This will be particularly important for a company that may have segregated its asset classes and used alternative methods for those respective classes."

Q&A

Risk talks to Kent Westerbeck, Chicago-based group senior vice-president at ABN Amro North America, about the sale of its mortgage servicing rights (MSR) portfolio to Citigroup, and the vagaries of fair-value accounting.

Risk: What immediate impact will the US mortgage business sale have?

Kent Westerbeck: A colleague here calculated that the firm had been spending an enormous amount of money in our adherence to hedge accounting for our former MSR portfolio. We were at a decided disadvantage compared to some of our peers: it was expensive and represented enormous accounting risk. With the MSR assets removed, a large portion of accounting infrastructure can be dismantled.

Risk: Is it true to say that accounting for derivatives seems to be in a state of flux in the US?

KW: Prior to the sale of our residential mortgage group in North America, ABN Amro North America was, in essence, faced with a tough choice: don't hedge MSR risk and so avoid the troublesome hedge accounting; or hedge and take on large accounting risk via IAS 39. Seeking hedge accounting had undoubtedly proven difficult for some banks in the US. We never had much trouble, but in the last year or so processes around hedging activities that were previously viewed as acceptable industry-wide appear to have been brought into question.

Risk: Can you give some examples?

KW: One example relates to immateriality. In the past, if you could demonstrate via an analysis that there was a risk that was minor, you would not be required to measure it. But now it seems immateriality may be becoming an inappropriate defence. Another evolving practice is around the designation of a hedge as fair value. To do this, you must, in effect, take your assets and break them up into homogeneous piles. Criteria for what constitutes homogeneity seem to be evolving. While the rules simply state you should have an expectation that assets are homogeneous, evolving practice is that back-testing must be used to establish homogeneity.

Risk: Accountants and auditors always want to try to stay ahead of current regulatory thinking in order to avoid the possibility of being required to restate. How have these apparent shifts been communicated?

KW: Regulatory staff meet industry professionals in informal discussion groups or make public statements that explicitly mention changes, or more subtly imply changes in thinking. These are then transformed into practice when accountants and auditors revise their interpretation of rules and advice.

Risk: Do these or other shifts have the potential to cause big problems?

KW: Perhaps the most worrying thing we have observed is the pushing by some of the dollar offset ratio test for hedge effectiveness for those that continue to want to achieve hedge accounting, without an appreciation of its flaws. The dollar offset ratio was an early approach used to test retrospectively whether a hedge had worked. The ratio in question is simply the change in value of the hedge to the change in value of the hedged item. If that ratio, as expressed as a decimal number, fell outside of the -0.80 to -1.25 range, then the test was deemed to have demonstrated hedge ineffectiveness. It was put forward as a logical test of hedge ineffectiveness, but was originally discredited in 1999. Research by Goldman Sachs demonstrated that the ratio had undesirable statistical properties and that even the most fabulous example of an economically effective hedge could fail the test nearly half the time. (See box: The upset of offset.)

Risk: So you favour regression or correlation-related tests?

KW: Yes, but accountants have suggested to us that it would be appropriate to use the offset ratio test as a double check for our statistical test. This would of course undercut the statistical test for no good reason, because as discussed it is fundamentally flawed and throws up false negatives.

Risk: Why do you think there's a lack of appreciation of the offset ratio test's weaknesses?

KW: It appears logical that the ratio should average around -1.00. The trouble is, as a statistic, the ratio does not have a well-defined mean or standard deviation, and it's this that people find very difficult to grasp. And when you start to explain using the calculus, you can see their eyes rolling to the back of their heads. It's a worry when an accountant insists that we should really back-test the evaluation of if a hedge was effective, and that they highly recommend it. You can pass the offset ratio test and that does not tell you it's a good hedge economically, and if it fails the test it does not tell you it's a bad hedge either. The test is worthless.

THE UPSET OF OFFSET

While the introduction of FAS 156 may have lessened the need for many mortgage servicers to seek true hedge accounting, the lure of hedge accounting remains for those firms that want to minimise earnings volatility. The dollar offset ratio test was, and is, viewed by most as a valid test of hedge effectiveness. Under the test, if the ratio of the change in market value of a hedged item to the change in market value of a hedge is between -0.80 and -1.25, the hedge is deemed effective.

In June 1999, Goldman Sachs sent a letter to the Financial Accounting Standards Board (FASB) in which it demonstrated that the dollar offset ratio was ill-suited to being used as a hedge effectiveness test. The 13-page letter proved that even the most highly correlated hedge could fail the test frequently. The analysis initially discredited the dollar offset ratio as a valid effectiveness test.

However, the dollar offset ratio test is still widely used to assess hedge effectiveness. Kent Westerbeck, Chicago-based group senior vice-president at ABN Amro North America, believes this is because many people still don't appreciate its flaws.

One of the main misapprehensions is that, for a good hedging relationship, if the dollar offset ratio was evaluated for a number of observations, many of the ratios would fall within the -0.80 to -1.25 range and would average at around -1.00, says Westerbeck.

Although Goldman Sachs highlighted that the ratio actually has no well-defined mean average, Westerbeck believes people don't accept this analysis. "Most people expect that the average of the ratio for a large number of observations - that is, the sample mean - will be around -1," he says. "They also believe that the statistic would follow the law of large numbers," he adds, referring to a piece of statistical reasoning that implies the sample mean of a statistic will, under certain conditions, converge to its mean as the number of observations grows larger.

"But the necessary conditions for this law to hold are not met by the dollar offset ratio statistic," explains Westerbeck. "Even for very good hedging relationships, the sample average dollar offset ratio does not tend toward -1.00 as the number of observations becomes large."

Experiments by Westerbeck and his colleagues involving hedges and hedged items that are 99% correlated hammer the point home. Randomly choosing 1 million samples of changes in hedged items and associated changes in hedges, they calculated dollar offset ratios for each. They then averaged 1 million samples of the ratio and repeated the process 1,000 times. The result was that the 1,000 average numbers mostly ranged from -10 to +10, with the averages frequently falling outside the -0.80 to -1.25 range that many intuitively expect.

The bottom line is that the dollar offset ratio test behaves differently to the way one would expect because its statistical characteristics are very different from the characteristics of statistics encountered in everyday life, says Westerbeck. "The dollar offset ratio test should not be expected to be between -0.80 and -1.25 when the hedge is effective," he cautions. "The offset ratio is an inherently inappropriate test of the effectiveness of a hedge, which is why we prefer statistical tests such as regression."

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact info@risk.net to find out more.

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here