The MBS market saw similar volatility in the third quarter to what was experienced during the second quarter, but the past three months tell a very different story than the preceding three. MBS of all coupons endured a pummeling through July and August with certain coupons losing >3.5% by early September. However, after the Fed declared they would not taper QE3, MBS and treasury rates made a complete turnaround and nearly all coupons of agency MBS finished up for the quarter. Mortgage REIT equities however weren't quite as lucky as most finished down for the quarter. From July 1st to September 30th the Market Vectors Mortgage REIT Income ETF (MORT) took a 4% hit, and industry leaders Annaly Capital Management (NLY) and American Capital Agency (AGNC) have endured 7.9% and 1.9% drops respectively.
The third quarter also gave investors some particularly notable events to digest regarding AGNC. In September management cut their dividend from $1.05/share to $0.80/share. This represents the largest dividend cut for AGNC in 4 years, and the first time in the company's history in which they have cut the dividend in sequential quarters. In an article during the second quarter, we assessed the impact of AGNCs positions and determined that based on their high volume exposure to assets in which depreciation losses would likely be realized as taxable losses, along with the need to restructure their portfolio in a volatile environment, AGNC would likely face headwinds in the taxable income category and that there was a significant probability the dividend would be cut to less and $1.00/share in the near term. In the discussion, we noted that distributing greater than $1.00/share in Q2 might leave management at risk of having to cut the dividend again in Q3. This assessment was validated by management in their Q2 earnings call as Gary Kain noted four weeks into Q3 that "Taxable income in the third quarter is currently biased materially lower given some of the actions we took in both the second quarter and third quarters, which will be realized for taxable earnings purposes in Q3." This was certainly a significant driver in the decision to cut the dividend again, but investors are still eagerly awaiting word on exactly where taxable income stands going into the fourth quarter of 2013. Simultaneous to the dividend announcement, management announced that they had spent $263M in order to repurchase shares of common stock at an average market value of $22.16, proving that at some point in the quarter this was a significant discount to book value. This is certainly a positive signal about the health of AGNC's portfolio in an unsettled environment, but investors are still unsure of exactly how well management protected book value through the duration of the roller coaster quarter.
With AGNC announcing their quarterly earnings and answering a few of those questions after hours Monday, we are going to break down what we know about AGNC's position based on their Q2 figures and the information management has made public in the past quarter; we will use this analysis to try and get a range on where we can expect book value to be for the earnings announcement, determine what other figures will be the focus of the release, and determine a proper approach to interpret their implications going forward.
MBS Portfolio Movement
Overall, the pricing delta on agency MBS looks favorable for AGNC, however there is a significant risk this quarter to taking the 'Rip Van Winkle' approach and looking at 6/30/13 versus 9/30/13 and calling it a day. As noted above, MBS prices dropped significantly intraquarter almost mimicking Q2 before a late September turnaround. As an illustration, the below graph represents the pricing of thirty year 4.0% coupon MBS (this is the highest volume coupon security in AGNC's portfolio as composed on 6/30/13).
The exact quantitative impact of this volatility is very hard to estimate with any accuracy given the large number of moving parts involved. Given the huge changes in each direction, management's particular repositioning actions and asset exposure on specific dates would be critical to the analysis and that information is not publicly available. The fact that swap prices performed effectively as hedges to MBS depreciation in Q3 is very important, and since AGNC was significantly hedged going into the third quarter, we assume that the extreme volatility does not invalidate the value provided by looking at the positioning of the portfolio as it was disclosed at the beginning of the quarter. We will assess all figures with the understanding that management of the portfolio in ways that significantly altered its exposure may cause material differences between the figures we derive and those realized in Q3.
To begin assessing how the quarterly movement affected AGNC, we will take a look at AGNCs MBS portfolio at the end of Q2 to get a feeling for how they were positioned going into Q3 and what rates were the most relevant.
If we look at page 6 of AGNCs Q2 presentation (above), we have the fair value of their MBS as of 6/30/13. If we map that to the market pricing on 9/30/13 we get the following changes.
Projected Fair Value
Projected Fair Value
*Holdings and Projected Fair Value Figures in Millions of $
In summing these figures, we derive that the portfolio AGNC held at the end of Q1 would have gained around $465M or $1.21/share. AGNC's quarterly disclosure included the mapping of their net TBA positions on the chart provided, which gives us a better capability to estimate the portfolio impacts from market value changes than was possible in previous quarters, so we will proceed with this rough figure rather than estimating an upper and lower bound.
Since we are only deriving one number and taking it as the best estimate, there is increased importance in qualitatively accounting for the additional information we have about the portfolio in Q3. As noted above, the relatively high volatility of the market with the scenario being a sequence of significant drops respectively followed a returns to the initial pricing results in a bias downward, as we assume AGNC is managing the portfolio with what is a relatively constant leverage and our model does not account for this effect. Basic bankroll management concepts that apply when assessing AGNC dictate that high volatility is bad for the long-term value of the bankroll, in this case AGNC's book value. In this explicit case, AGNC's starting exposure for the quarter will naturally decrease as book value drops due to MBS depreciation. After adjusting for the losses and maintaining a constant leverage, the portfolio would have less than the initial exposure for the subsequent rebound, and the pricing adjustment would have to bypass the starting pricing before AGNC's portfolio matched its initial value. In this case, the gains on hedges will help to mitigate this effect, as hedges increased in value and then would presumably be scaled down rather than up with the slight losses to the MBS portfolio, however this concept still applies even if it is damped. On the positive end, if we assess management's approach to positioning the portfolio through the second quarter, we find that continued implementation of that strategy is significantly beneficial for the portfolio's value. Among a few other optimistic trends, management increased holdings in 15-year MBS from a 34% composition of the portfolio to 42% going into Q3 and in the second quarter earnings call, Gary Kain noted that this percentage had continued to creep upwards and was higher than the 42% quoted. These 15-year securities performed better than their 30-year counterparts and represented a much less volatile investment when compared to the 30-year securities. We will proceed with the rough estimate derived above while keeping these concepts in mind.
As of 6/30/13 the net spread on AGNC's portfolio including TBA positions was 1.59%, which dropped from 1.71% on 3/31/13 as mortgage rates rose during Q2 as an artifact of increasing hedge costs. As rates continued to rise for the majority of Q3, it is very probable that AGNC's CPR will come in lower than projected. Also notable is that as of 6/30/13, notional value of hedges covered >100% of AGNC's MBS assets so it isn't probable that the hedge/MBS ratio will increase further. Any transitions to more expensive hedges such as swaptions will surely weigh on these figures, however the swap and swaption segments of the portfolio raised slightly in Q2 while the less expensive treasury hedges were contracted. Given that those adjustments have already taken place to a significant extent, it isn't likely that the spread would shrink much further based on hedge composition as management would effectively be phasing out direct treasury positions if this were the case. Although the final numbers for the quarter showed a slight rate contraction, MBS rates were up for the majority of the quarter and it can be assumed that the asset yield on the portfolio was higher through the quarter than the figures reported on 6/30/13 would suggest given management's repositioning actions, even if only slightly.
Beyond simply determining their portfolio spread, it is relevant to recognize that significant book value changes will cause management to alter the size of their portfolio to maintain the same leverage. AGNC was very well hedged going into Q3 and the portfolio was not extremely sensitive to rate changes, but changes through two thirds of Q3 were very similar to the figures realized in Q2 that caused AGNC to contract its MBS holdings by >10%. The market value of the average holdings during Q3 was almost certainly lower than the slightly >$90M reported on 6/30/13.
Based on our derivations, let's place our net spread projections between 1.6% and 1.8% and the average portfolio value between $85B and $90B. Those figures would mean AGNC has earned between $340M and $405M or between $0.88 and $1.05 in interest income per share.
In breaking down AGNC's hedge performance in Q3, It should first be noted that short-term hedge assets (treasuries and swaps) held for the duration of the quarter lost value while longer-term assets appreciated. Ten-year treasury yields rose 13 Bps in the third quarter from 2.49% to 2.62%, while two-year treasuries fell 3 Bps from .36% to .33% (five-year treasuries finished unchanged). Given that fact, the composition of their swap portfolio (shown below) suggests approximate performance parity across their holdings in this area, as the average years to maturity is just over 5 years. For the purposes of this snapshot analysis, we will approximate that the swap portfolio was roughly unchanged through Q3.
The composition of their swaption portfolio is a bit more weighted toward longer swap terms (8.4 years) than the swap portfolio. Extrapolating that on average, the related move for underlying swaps related to the swaptions moved up about 10 basis points, and applying the average duration of 3.9-year average duration AGNC quoted on their swaption holdings, we derive an estimate of $93M in gains on swaptions (not including time depreciation of the swaptions).
Lastly, AGNC's net treasury positions were aligned very well as they had small positive bets in the 5-year while realizing gains on negative positions on the 7-year and 10-year treasury. If we aggregate the 6/30/13 figures, this projects to ~$70M in additional gains.
So in aggregate, we estimate that AGNC has seen ~$163M or 0.42/share in gains from hedging activities. After deriving all of this, it's vital to note that the true figures for Q3 will almost certainly be more positive than the snapshot derivation. Throughout the quarter, Treasuries were consistently higher than the opening or closing values, with the 10-year breaking the 3.00% barrier and reaching a high of 52 basis points greater than the 6/30/13 numbers (to provide some context, the quarterly rise in Q2 was 64 basis points). While the value of the MBS portfolio dropped and the hedge portfolio rose, management almost certainly realized gains on hedges to some extent and liquidated some swap, swaption, or treasury assets at a higher value than quoted as of 9/30/13 above. It can't be determined which assets were liquidated, so there isn't a strong basis to estimate a quantitative benefit to the assumed action, but as noted above we will proceed with the understanding that the underestimate of the hedge performance will roughly offset the variance based overestimate of appreciation on the MBS portfolio.
So What is Our Book Value Estimate and What Does it Mean?
If we apply our derived variables above, we arrive at the following figures for quarterly gains on book value through Q3:
MBS appreciation + Hedge appreciation + Interest Income + Buyback benefit - Dividend Payout
$465M + $163M + $372.5M + 41M - 307M = +734.5M
Those gains would represent a very strong book value performance for the quarter of ~$1.91/Share in appreciation, and a new approximate NAV of $27.42
It is important to remember that there is a good deal of uncertainty to this estimate. What we can determine with relative confidence is that AGNC is likely still trading at a very significant discount to its Q3 book value with market value closing at $24.05 on the last trading day before earnings, and a market price drop would likely be supported by further buybacks.
The most important figure for the dividend is the taxable income figure, which only includes the realized items on the balance sheet. This includes interest income, any closed out MBS positions, and any expired or closed out hedge positions. For Q3 our projections suggest between $0.88/share and $1.05/share in interest income, but the other items remain a mystery.
There is no basis to project with any confidence what total value of the losses AGNC will realize in order to reposition given the current environment is, or which TBA positions they will choose to close out and realize losses on versus take delivery of the assets and hold them. What we do understand is that very late in the quarter, management judged that the appropriate dividend distribution was $307M or $0.80/share and that there was $425M or $1.07/share on the coffers on June 30th. In iterating through AGNC's recent quarterly figures, it is apparent that they were carrying huge basis losses on the assets in their MBS portfolio, potentially in the billions and certainly greater than the interest income that was earned this period. Also, AGNC management knew at the beginning of Q3 that the taxable income figure was already materially biased downward based on significant repositioning action. Given this, there is a very large range of possible scenarios for taxable income in Q3, including a very real possibility for a negative figure to be reported.
The taxable income uncertainty is likely a significant driver to the deep market discount relative to book value. Taxable income will almost definitely be the most important variable in the upcoming announcement and the market reaction will key off of this figure; the undistributed taxable net income held by AGNC will provide investors with guidance as to whether the $0.80 dividend was likely a one-time low distribution needed to protect book value in a volatile environment, or a signal that AGNC has accrued significant losses and will be spending multiple quarters rebuilding their taxable earnings to support their dividend. If our projections are correct, our book value estimation leads us to believe that there is likely minimal downside risk to a rational negative earnings shock in the area of taxable earnings, and that a significant market drop would likely be quickly supported by management buybacks. In a positive scenario for the taxable income figure, AGNC's discount to book value leaves significant room for market appreciation in the short run and we feel comfortable holding an exposed long AGNC position through Q3 earnings despite the high level of uncertainty regarding the taxable income figures.