US Banks: Duration and Convexity Risk in AFS Debt Portfolios
SEE LAST PAGE OF THIS REPORT Howard Mason
FOR IMPORTANT DISCLOSURES 203.901.1635
July 1, 2013
US Banks: Duration and Convexity Risk in AFS Debt Portfolios
- Given the surge in long rates, we expect Q2 unrealized losses on available-for-sale or “AFS” debt securities at US commercial banks to approach total earnings and prevent meaningful growth in book value.
- Our forecast is for losses of $25-30bn on AFS portfolios of ~$2.5tn compared with total bank earnings of ~$40bn; the 10-year Treasury yield increased ~0.6% in the quarter to 2.5% versus a 0.1% increase in Q1 when AFS losses amounted to $6bn.
- We note that unrealized gains/losses on AFS portfolios flow to equity through other comprehensive income, not earnings.
- Some regional banks with large holdings of residential mortgage-backed securities, such as STI, BBT, RF, and FITB, will likely see flat-to-slightly-down book value (see Exhibit below) with a more meaningful decline in book value at BAC of ~$3bn or ~25-30 cents/share.
- Our loss estimates may be conservative given the “negative convexity” of MBS portfolios: duration extends as rising rates slow prepayments.
- At BAC, in addition to AFS losses of ~$6bn from debt securities, we expect a loss of ~$150m from an AFS-designated equity stake in China Construction Bank.
- We do not expect hedging activity to meaningfully affect our conclusions given AFS portfolios are themselves hedges against the impact of changing interest rates on the economic value of bank deposit franchises.
- It is an asymmetry of bank accounting that losses on the AFS portfolios are taken immediately to equity while gains on the deposit book are recognized to equity over time through higher net interest income.
Exhibit: After-tax AFS losses versus Earnings at Selected US Banks (ranked by estimated duration of AFS book)
Given the recent surge in long rates (see Exhibit 1), investors are concerned about mark-to-market losses in bank portfolios of debt securities, particularly mortgage-backed securities. Typically, these portfolios carry an accounting designation of “available for sale” or “AFS” so that unrealized gains and losses flow to equity through other comprehensive income; realized gains and losses flow through earnings, of course.
Exhibit 1: Recent Changes in Treasury Yield Curve
Source: Federal Reserve
Regardless of accounting designation, there is the potential for debt securities held at fair value to generate large hits to equity in Q2 potentially delaying plans to return capital to shareholders. This is widely recognized in the industry with a senior executive of a top US bank commenting recently to the Financial Times: “I would think most institutions are going to have a fairly sizeable hit to their equity. You’ve really had this concentrated one or two week period where all hell is breaking loose”.
Estimating Aggregate Losses for US Commercial Banks
The possible extent of negative marks is revealed by Government data (through the H8 release from the Fed Reserve) indicating that in the month of June alone (through June 19th) US commercial banks had unrealized losses of ~$17bn. We obtain this amount as the difference between the accumulated unrealized gain of $33.2bn in March and the corresponding figure of $14.8bn for the week of June 19th (see Exhibit 2). We note these figures are all after tax.
This loss of $17.4bn represents 0.7% of the fair value of AFS portfolios of ~$2.5tn and compares with an increase in the 10-year Treasury yield over the period ~0.4% to 2.3%; in other words, the (modified) duration of the fair value of bank AFS portfolios appears to be just shy of 2x with respect to the 10-year Treasury yield. This duration estimate is consistent, albeit on the possibly conservative side, with the $6.5bn decline in AFS portfolios for Q1 which represented 0.26% of fair value versus an increase in the 10-year Treasury yield of 0.1%.
Assuming AFS portfolios are valued consistent with an end-quarter yield of 2.5% on the 10-year Treasury, the Q2 increase in this yield is 60 basis points. Given our duration estimate, this generates an (after-tax) loss estimate for the portfolios of 1.1% of fair value or $25-30 billion.
Exhibit 2: Accumulated Unrealized Gains on AFS Portfolios at US Commercial Banks in $billion
Source: Federal Reserve. Data are averages of values on Wednesday of each period
Estimating Losses for Selected US Banks
Exhibit 3 shows SSR estimates for the duration of the fair value of AFS portfolios at selected US banks with respect to the 10-year Treasury yield. The average duration is just shy of 2x since the ~0.1% increase in this yield in Q1 to ~1.9% led to an average decline in the carrying value of AFS portfolios of just under 0.2%; this is slightly lower than the result of 0.26% for US commercial banks in aggregate largely because of the impact of C which had a Q1 gain on its AFS portfolio (because, uniquely among the banks in the sample, over one-third of its AFS portfolio is in foreign securities).
BAC is another outlier with an AFS portfolio decline in Q1 of ~0.4%. During the quarter, the accumulated pre-tax gain on AFS debt securities declined by $1.5 billion from $7 billion to $5.5 billion; the after-tax decline was $1.1 billion. The bank also designates as AFS some marketable equity securities, almost entirely consisting of its holding of 2 billon shares, or ~1%, of China Construction Bank (“CCB”) but these produced no unrealized gain or loss in the quarter. Outliers on the other extreme (i.e. with below-average durations) are PNC and KEY.
Exhibit 3: Estimated Duration for Fair-Value of AFS Portfolios at Selected US Banks
Source: SNL, FRY9-C, SSR Analysis
In Exhibit 4, we estimate Q2 losses by combining the duration estimates for individual banks from the Q1 results shown above with an assumption that AFS portfolios at end-Q2 will be valued consistent with a 10-year Treasury yield of 2.5% (representing a 0.6% increase for the quarter). It is clear that some banks will show losses that are large in absolute terms and relative to Q2 earnings. In particular, it is unlikely there will be meaningful growth in book value at banks with relatively high-duration portfolios including STI, BBT, RF, and FITB; and, at BAC, we expect a loss of book value of ~$3 billion representing 25-30 cents/share.
Exhibit 4: After-tax AFS losses vs. Earnings (ranked by estimated duration of AFS book)
Source: SNL, SSR Estimates
In practice, duration is an incomplete measure of the sensitivity of AFS fair values to changes in interest rates. This is because large portions of the AFS portfolio (~55% for all US commercial banks and 80%+ for BAC, STI, BBT, and RF – see Exhibit 5) are mortgage-backed securities or “MBS” which have so-called “negative convexity”: their duration increases as interest rates rise because borrowers have less incentive to prepay and refinance.
Note: This decline in “prepayment speeds”, and resulting increase in duration of the mortgage-backed securities, is sometimes referred to as “extension risk”. Commercial MBS or “CMBS” tend to have less extension risk than residential MBS or “RMBS” because lenders typically obtain contractual prepayment protection.
Exhibit 5: Proportion of AFS Portfolios in Residential and Commercial MBS
Source: SNL, SSR Analysis
Exhibit 6 shows that our duration estimates for the fair value of AFS portfolios in Exhibit 3 correlate quite well with the weighting of residential MBS or “RMBS” within the portfolio in Exhibit 4. It is likely that this is because as interest rates rose in Q1 2013, those banks with higher weightings of RMBS saw their AFS portfolios extend through the quarter raising the average duration for the period. With the greater rise in interest rates in Q2 2013, we can expect the duration to increase again so that our loss estimates in Exhibit 3 could prove conservative particularly for those banks with relatively large RMBS books.
We do not include C in the Exhibit because of its exceptional foreign debt holdings as discussed above. Among the banks which are included, KEY, PNC, and MTB are outliers in that the duration estimate is lower than would be expected from the size of their RMBS books; MTB, in particular, appears to have an AFS portfolio that is entirely comprised of RMBS with no US Treasury or agency securities. BAC and STI have large RMBS books and high duration estimates for the size of these books.
Exhibit 6: Estimated Duration of AFS Portfolio (y-axis) versus Proportion of RMBS (x-axis)
Risks to BAC Analysis
Forecasting changes in bank balance sheets is obviously fraught: they are complex and the accounting rules can be counter-intuitive; there are many moving parts (including hedging activity); and we have available only highly-aggregated position data. However, we believe our estimate for the AFS and debt securities losses at BAC is more likely to be conservative than aggressive for the following reasons:
- 82% of the AFS portfolio comprises RMBS securities and, given extension risk particularly in June when rates moved up sharply, we may have under-estimated the average duration of the portfolio through the quarter. For example, if the duration is 5x rather than our estimate of 4x, the loss estimate increases by $1.5 billion.
- We expect a loss in the AFS portfolio of an additional $150m from BAC’s equity stake in China Construction Bank. At end Q1, this had a fair value of $1.5bn and shares have traded down ~10% in the quarter.
- Unlike peers, BAC has a large holding of debt securities (aside from the trading book) held on a mark-to-market, rather than AFS, basis. Specifically, these securities amount to ~$50 billion and could generate losses of an additional ~$1bn. These losses, unlike those on AFS debt securities, flow to equity through earnings (and, specifically, through the “other gains/losses” component of non-interest income).
We do not expect hedging activity to meaningfully affect our conclusions, and note that in Q1 the loss on the AFS portfolio, excluding derivatives, was ~$1.1 billion while the loss, including derivatives, was $1.3 billion. The reason is that the AFS portfolio is already a hedge against the impact of changing interest rates on the economic value of a bank’s deposit franchise. It is one of the asymmetries of bank accounting that in a rising rate environment the loss on the AFS portfolio is taken immediately to equity while the gain on the
underlying position, in the case the deposit book, is recognized to equity over time through higher net interest income.