I have been attracted to American Capital Agency (AGNC) by the high dividends since 2010 and the handling of the position had been straightforward until May 2013 (buy and buy more). Things changed in May 2013 with the book value hit. Instinct kicked in, I did some sales but I missed out on the balance of sales by being penny foolish.
So now with a reduced holding and better emotional stability, I decided it was time for another look at AGNC but from a different perspective.
With the ongoing taper talk, it made sense for AGNC's management to run a more duration neutral portfolio, as per their June 2013, 10Q.
The purpose of this write-up would be to look at AGNC in a duration hedged perspective and the possible drivers of risk to book value and implicitly the level and sustainability of dividends.
MBS can be decomposed into "Interest Rate Risk" + "Mortgage Spread Risk".
Mortgage spread risk can further be broken down into credit risk (which I will assume to be negligible since agency MBS) and prepayment risk (within which interest rate volatility being a component).
Since management is now running a duration neutral book, most of the interest rate risk has been stripped out (for small changes in interest rate) and the portfolio is now left with mortgage spread risk.
Dividends in a duration hedged portfolio but under different hedging scenarios.
As of June 30 from Q2 2013 10Q, AGNC had the following MBS holdings and the implied calculated annualized cash flows as shown in table (1):
Table (1): MBS holdings and annualized cash flows
Case (1): Duration hedged, no swaptions and assume same notional required for swaps
In case (1), I have used the metrics provided by the 10Q of Q2 2013 and have made one assumption regarding the swap notional required for a duration hedged portfolio, in this case assumed to be $77,000 million.
Under this scenario, the maximum dividend can be around $2.42 per year, around 10.7% yield.
One thing worth pointing out is that even if the Fed hikes rates aggressively and yield curve eventually goes flat to say 5% across maturities, assuming mortgage spreads remain the same, the above analysis still holds true.
Repo (funding) cost goes up but so does the float side of swaps (Libor), and as long as the funding spread remains 0.18% (0.44% - 0.26%), we can expect same cash flows.
What matters most in a duration hedged portfolio is the MBS spread (MBS and TBA Yield - Swap Hedge Fixed rate) vs. the Libor repo spread (Repo - Libor on float side of swap).
In terms of the assumption regarding the use of $77bln of swap notional, my suspicion is that
(i) either the notional is too small or
(ii) the yield of the hedge swaps on the net will be higher,
so either way the bias points to a lower dividend.
Case (2) reducing swap notional and adding swaptions
Case (2) is more representative of AGNC's holdings as of Q2 2013 where management used swaptions and swaps.
Even though there is no recurring cash flow on swaptions, there is time decay on the swaptions and a best guess estimate would put the time decay to around $200mm per year. This marginally improves the dividend to $2.64 per year, around 11.7% yield.
Both case (1) and (2) imply that under current MBS and funding spreads, the dividend should be sustainable in the 10-11% yield region.
Now 10-11% dividend is still better than most dividend yields out there, but we need to evaluate what kind of volatility to book value is to be expected to "earn" this 10-11% dividend yield.
In other words, a duration hedged portfolio should bring stability to book value but how stable is book value?
What can drive mortgage spreads wider and by how much?
(1) Some risks affecting CPR are somewhat unquantifiable - HARP and other forms of government intervention, ongoing gradual phasing out of Fannie and Freddie with possibility of being replaced by another agency.
(2) Rate volatility leads to duration mismatch - the only way to cleanly hedge is to use swaptions but this is not free.
The Q2 2013, 10Q provides a good interest rate sensitivity as shown in table (4):
Table (4): Interest rate sensitivity
A 100bp shock higher in rates seems to "only" lead to a -5.95% loss in book value which is more palatable from the recent episode from May 2013, honestly not a disaster.
(3) MBS spread widening - how much can MBS underperform relative to the hedges and its impact on portfolio?
Charts (1) and (2) below show the history of 30yr MBS spreads to 10 year USD swaps.
Caveat - Although this spread is not exactly reflective of AGNC current portfolio spread, it can give a sense of the volatility of spreads and where we are relative to history.
Chart (1): History since 1988
Chart (2): History since 2008
MBS Spreads - Various Statistics over 2 periods:
Table (5): Statistics of Spread
Since 2008, there has been a higher equilibrium (average) for MBS spreads relative to the 1998-2007 period, possibly from phasing out of Fannie and Freddie.
A 1 standard deviation in MBS spread is around 31-33bp and an approximation of a 30bp widening is given in table below:
(Duration * notional* basis point widening would be a good proxy for $ hit for small moves in spreads - Ignoring some negative convexity impact here).
Table (6): Book Value sensitivity to MBS spread
Assuming a duration of 7.5, table (6) indicates that if MBS spreads widen out by 30bp, we can expect a book value hit of $5.20.
Market Update - Note that 30 yr MBS spreads as of 8/29/2013 have tightened by 23bp from where they were on 6/27/2013 (only have weekly data so no data point for 6/30/2013).
Using above proxy calculation, new book value could be around $28.05 - $28.85 as shown in table (7): (reiterating the caveat that I have used 30yr MBS spreads to proxy for AGNC current holdings which they have shortened in maturities).
It can be argued that although a widening of spreads is a Mark-To-Market issue, the problem is that the company still needs to satisfy margin calls and the degree of leverage employed will impact its ability. The worst case would be for management to have to sell MBS to satisfy margin calls which crystallizes losses and reduces book value.
Else management could resort to issuing more shares to raise cash, but this will not be fast enough and does not help existing shareholders.
Ideally in times of widening you would want management to use unencumbered cash to add to leverage and hope that spreads tighten. But as always easier said than done.
(4) Funding spreads widen out and becomes less available - this reduces leverage and future dividend flows.
I cannot help but think that this is as good as it gets on that front with haircut no more than 5% and repo around Libor +13bp. A bigger haircut will not necessarily impair book value directly but this would imply a lower leverage and hence a lower dividend yield.
Conclusion - Under a Duration Hedged Portfolio
- It seems to me that an 8.5x leverage in a duration hedged portfolio can yield around 10% with the potential hit from a 1 sigma move in MBS spreads to be worth around $5 on the share price.
- A flat yield curve will not necessarily imply a reduction in dividends from the 10% level.
- Since portfolio is closer to being duration hedged, the better indicator for AGNC would be MBS spreads and to a lesser degree funding - Libor spread.
- Market's perception of consistently repricing AGNC to outright moves in rates seems misplaced to me at the moment.
- Finally, the good news is that AGNC is trading to a decent discount (1 sigma MBS spread) to approximated book value.
- Somewhat subjective, but in the short term, if the dividend yield does decline to around 10%, I am concerned that market's singular focus on dividend yield may lead to AGNC trading even cheaper to book value. So to me the risk on AGNC at the moment is not necessarily increasing interest rates but more market's habit of viewing AGNC with at least 14% dividend yield.