- The key investment issue – Can AGNC maintain its dividend?
- Managing credit risk
- Managing interest rate risk
- Managing duration risk.
- Managing prepayment risk
- Managing leverage risk.
- Can AGNC’s operating earnings cover its dividend? It has so far.
- My earnings model
- Why I don’t think interest rates will increase a lot.
- We’ve made it! Estimating AGNC’s ’23 EPS
- Net/net, the odds are good that AGNC can maintain its dividend. I’d buy some. Beats the heck out of a 2% Treasury bond yield.
The key investment issue – Can AGNC maintain its dividend?
AGNC has paid a $1.44 annual dividend for the past two years. At its current price of $14, that’s a 10.3% dividend yield. As a mortgage REIT, we have to assume little or no long-term growth in the dividend. But if AGNC can just maintain it, the resulting 10% annual return would be pretty sweet, very likely outperforming the S&P 500 total return going forward.
Can AGNC maintain its dividend? I believe it will for the foreseeable future. I’ll start from the beginning of the story to convince you of that claim.
What AGNC is – a short version
AGNC is a specialized investment company. Its specialty is “agency” mortgage-backed securities (MBS). The “agencies” are Fannie Mae and Freddie Mac MBS. AGNC finances the bulk of its MBS with “repos”, which are bank borrowings that use the MBS as collateral. The key to understanding AGNC is understanding the three risks this business model is subject to:
- Credit risk
- Interest rate risk
- Leverage risk
Managing credit risk
This one is easy. The mortgages in an agency MBS are insured by Fannie Mae or Freddie Mac. If one of the mortgages in the MBS defaults, Fannie or Freddie continues making the principal and interest payments to the MBS investor. AGNC has a small portfolio of non-agency MBS investments, small enough to ignore. So no credit risk.
Managing interest rate risk
This is the big one for AGNC. AGNC manages two major interest rate risks – “duration” risk and “prepayment” risk. Let’s take them one by one.
Managing duration risk.
AGNC’s agency MBS are nearly all fixed rate mortgages, mostly 30-year with some 15-year. A newly originated 30-year MBS has, say, an expected 6-year life (people refinance or move), or “duration”. But the repos with which AGNC finances the MBS have about three month lives. Let’s say the MBS yields 3% and the repo yields 0.25%. Fine so far. But if the repo yields rise appreciably over time, AGNC’s interest income will get squeezed or even turn negative.
AGNC manages duration risk by buying hedges that extend the duration of its repo debt. For example, at the end of 2021, the average maturity of its repos was 63 days. But the hedges it bought extended the overall duration of its debt to 4.0 years, nearly the same estimated duration of its mortgages at that time.
This chart shows a history of AGNC’s hedging activity. The “net duration gap” is the difference, in years, between AGNC’s estimated asset duration and its hedged liability duration. The “hedge ratio” is the percent of its debt that it has hedged.
Source: AGNC financial reports
You can see that AGNC actively manages its hedge position based on its judgment of the direction of interest rates and how the expected rate moves will impact the income and risks of its MBS portfolio.
Managing prepayment risk
AGNC nearly always buys MBS at a premium above par; generally 3-4 points over a 100 par value. But borrowers will only pay back the 100. AGNC knows this of course. Its pricing assumes that the 3-point premium over par will be written off over the, say, 6-year expected life at 0.50% a year. But what if interest rates decline sharply and borrowers start actively refinancing and the 6-year life now becomes two years? Then that premium AGNC paid has to be written off over only two years, costing it 1.50% a year. That would gobble up the great bulk of AGNC’s earnings on that investment.
AGNC manages prepayment risk in two ways. One is by recognizing prepayment risks in structuring its hedges. The other is by buying MBS it expects to refinance at a below-average rate. An example is a small-balance loan (say $40,000) where the financial savings from refinancing is modest. The result is that AGNC’s percent of loans refinancing has consistently been well below average:
Managing leverage risk.
The less investor equity capital that AGNC uses to finance its MBS assets, the greater the equity investor return. But also the greater the risk of insolvency if those duration or prepayment risks go seriously wrong. Here is a history of AGNC’s financial leverage, measured as its MBS assets divided by its shareholders’ equity:
Source: AGNC financial reports
This chart has two highlights:
- The leverage range over time has been low. The 6-10 range compares to 10+ ratios for banks.
- The current reading of 7.7 is below average. As AGNC said on its Q4 earnings conference call, “We can comfortably operate at 8.5, 9.5 leverage…I think we will over the next couple of quarters be able to take advantage of this environment once the [bond] market finishes its [Federal Reserve] repricing….”
Can AGNC’s operating earnings cover its dividend? It has so far.
First, an annoying but necessary accounting digression. I and Wall Street use what AGNC calls its “net spread and dollar roll income per share” as operating income, rather than its reported GAAP income. For some reason, the accounting gods in their infinite wisdom use a mix of accrual and mark-to-market accounting for companies like AGNC. Trust me, you don’t want to know the gory details. But the fact is that AGNC’s operating income has driven the dividend, not reported income, as this chart shows:
Source: AGNC financial reports
AGNC cut its dividend in early 2020 in response to the uncertainty of COVID’s impact on Federal Reserve interest rate policy. Clearly it turned out to manage those policy changes quite well. The Fed is now of course radically pivoting from massive easing to aggressive tightening. What does that mean for AGNC’s operating earnings going forward?
If you believe Wall Street, AGNC has things covered. Wall Street analysts, as reported by Seeking Alpha, sees ’22 operating EPS coming in at $2.49 and ’23 at $2.32. But you’re paying me the big bucks for my view on earnings. Here it comes:
My earnings model
AGNC’s operating income has remarkably few moving parts. I’ll use Q4 ’21 as the example:
- +$440 million of net interest income
- ($20) million of operating expenses
- ($25) million of preferred stock dividend expense
- Equals $$395 million of operating income
Interest income in turn is a function of AGNC’s average MBS assets times its net interest spread. For Q4, those numbers were $82.6 billion of MBS assets times a 2.15% net interest spread.
The big swing item is the interest spread. What drives it? After some math that frankly should put me in line for a Nobel prize, I came up with a formula that explained 85% of the change in AGNC’s interest spread since 2015. It is:
+ .009 x AGNC’s “TBA” share of earning assets
+ .029 x AGNC’s loan repayment rate
– .126 x the fed funds rate
“TBA” stands for “to be issued”. They are newly originated MBS. TBA’s for some complex reasons get “special” when the Fed is actively buying MBS. The industry term “special” means “more profitable”. So when the Fed is active in the market, like it has been recently, AGNC has more TBA assets and earns more interest income.
Next, the loan repayment rate is the percent of AGNC’s MBS that pay down due to refinancing. Because AGNC has done such a good job reducing its prepayment risk, it is more profitable when homeowners actively refinance.
The fed funds rate is, well, the fed funds rate. AGNC’s repo financing prices off of fed funds, so the higher fed funds goes, the higher AGNC’s borrowing costs, even after hedging.
These three variables are largely driven by Federal Reserve actions. So what will the Fed do over the next few years?
Why I don’t think interest rates will increase a lot.
We all know that the Fed is raising rates. The issue for AGNC (and for lots of others!) is “How high?”. My belief is “Not that high”. The reason is this number – $65 trillion. That is how much total debt – consumer, business and government – the U.S. currently has. Federal government debt alone is $25 billion. Total debt now is about 3 times GDP, up from the 1.3 times it was in 1980, the last time the Federal Reserve had to deal with high inflation.
Here’s the issue with the $65 trillion. Let’s say the Fed raises the whole yield curve by two percentage points. That extra 2% increases the federal Government’s interest payments, and the deficit, by $500 billion a year. That means either tax increases just to cover at least part of the cost, or the inability to add social programs or other spending Congress is considering. Big problem.
As a result, I think that part of the Federal Reserve’s mandate going forward is to help manage the country’s debt load by keeping interest rates as low as practicable, considering inflation.
We’ve made it! Estimating AGNC’s ’23 EPS
I now apply a reasonable range of Fed policy paths to my earnings model. Let’s start with the historic range of the three variables in my earnings model and how I think they will change for three Fed policy paths:
Sources: My estimates and AGNC financial reports
“Easy” Fed policy is modest rates increases, “tight” is large increases.
Here is how these variable ranges drive my interest spread model…
…And then my 2023 operating EPS forecast:
My “easy” Fed policy forecast assumes that fed funds reaches 2% in ’23. Interestingly, that is the Fed’s current forecast. The latest estimates for Morgan Stanley and UBS are for 1.5% and 2.0% respectively. If so, AGNC can comfortably maintain its dividend. Only in my “tight” policy outcome will AGNC have to cut its dividend, and then only to $1.00, or a 7% yield. And the 4% fed funds rate will clearly seriously create material debt service problems for the government and the private sector, so I believe the Fed will desperately try to avoid it.