In my previous article I looked at how the dividends of Annaly Capital Management (NLY) varied as a function of the spread between the yields on the 20 years and 3 months Treasuries, going back to October 1997. I chose Annaly because it is the largest mortgage Real Estate Investment Trust [mREIT] in the US, is considered a bellwether of the industry, and its 10Q files (quarterly reports) go back more than a decade. The purpose was to show that, historically, a wide interest rate spread was correlated with high dividends.
Historical perspective is somewhat biased ("past results are no guarantee of future performance"), especially considering that since its inception Annaly had to deal with an environment of mostly decreasing long-term interest rates, while most people now believe that long-term rates are mostly going to rise. However, the basic business model of Annaly is not incumbent upon the direction in which interest rates vary, and therefore it is likely that past behavior can be used as a guide for what to expect in the future. Theoretically, if the management team of Annaly properly hedges against rising long-term rates (through interest rate swaps), they should be able to continue taking advantage of a broadening yield spread as they did in the past: a higher yield spread means a higher interest income, and due to the REIT status of NLY (which forces the company to pass 90% of its taxable income to shareholders) this translates into a higher dividend.
I received several comments regarding the fact that I only considered the yield spread as a source of income for NLY, ignoring the impact of other quantities and especially a change in fair value of mortgage-backed securities [MBS]. It was pointed out that a change in the book value [BV] of Annaly also impacts its earnings (through capital gains or losses) and its capacity to maintain the dividend. This is very relevant as the interest rate environment since May 2013 deeply affects the BV of mREITS: the quick rise in long-term rates produced a drop in the value of Annaly's portfolio of MBS. By historical standards, the speed at which this occurred was quite unusual.
A quick drop in the BV of Annaly has immediate negative consequences: for instance mREITs are forced to sell some MBS (at a loss most likely) to maintain their leverage ratio (as reported here), and if there have been any margin calls from the financial institutions that Annaly borrowed money from (through repurchase agreements), then more MBS had to be sold. It is almost certain that for the 2nd quarter of 2013, Annaly will realize a significant capital loss and record a large drop in BV. Will that affect the next quarter dividend? Annaly already cut it in early June, in response to the change in interest rates environment. MREITs may reduce their dividend for the following non-exhaustive reasons: the forced sale of part of their portfolio negatively impacts their interest income, and/or their book value drops so much that they may want to keep some cash to prevent further loss in BV (the realized capital loss, lowering their taxable income, allows them to hoard this money rather than distributing it), and/or they may want to use part of that cash to reimburse some of their debt.
Therefore, there are two conflicting tendencies when long-term rates rise (and short-term rates don't change): a drop in BV potentially accompanied by a dividend cut, and a rise in yield spread potentially generating a dividend raise. Here I look at past results of Annaly to see how yield spread, book value, and two other parameters impact the dividend. I try to determine which factor, amongst the few I consider, has historically been the best predictor of Annaly's quarterly dividend, and what we can expect for the upcoming quarters.
Dividends and correlation coefficients r
In this section, correlation coefficients between dividend yields of Annaly and a few select parameters are computed. First, a word of caution: the Pearson (linear) correlation coefficient r is a quick and convenient way of determining the strength of a relationship between two parameters, but it only tests for a linear relationship and ignores more complicated ones. Moreover the value of r can be significantly impacted by the presence of a few outliers. The sample size of the parameters I use is rather small (only 45 quarterly results for NLY, from 1st quarter of 2002 onwards), and therefore the r computed have significant uncertainties. For instance, for r=0.75, the standard deviation (which can be interpreted as the error bar, or uncertainty on this measurement) is 0.065, i.e. a relative uncertainty of about 8.7%. Finally, there is the well known adage "correlation is not causation": two variables may be strongly correlated but it does not necessarily imply that one variable is the cause of the other.
First is a figure similar to Figure 2 of my previous article: it shows the dividend yield of Annaly as a function of net interest spread (the dividend yield is calculated using the share price at closure of the last trading day of the quarter). As in the previous article, I used the 20 years- 3 months Treasuries yield spread. I looked at 44 quarters, ignoring 1 quarter: the 3rd quarter of 2003 which is a clear outlier, due to its abnormally high constant prepayment rate of 48% that is extremely unlikely to happen again anytime soon.
For these 44 quarters, the correlation coefficient between dividend yield and Treasuries rate spread is r=0.875. A potential bias of the analysis performed for my previous article is that I mainly compared dividends and interest rate spreads for the same quarters. However, if the spread changes during a specific quarter, it might not be reflected in the dividend until several quarters later. Therefore the values of the correlation r between yield spread and dividend yields were also computed for the same quarter, and for 1, 2, and 3 quarters later, respectively: r=0.825, 0.794, 0.693, and 0.548 (for simplicity, I used all 45 quarters when computing these numbers, including the outlier. You can see how its presence lowered r from 0.875 to 0.825 for same-quarter variables).
The differences between all of these values are within the uncertainty on r, but it seems that maximum correlation is reached for a zero quarter difference: on average, the rate spread at a specific quarter impacts the dividend at this same quarter or the next. The rate spread is quickly reflected in the dividend. Incidentally, the net interest margin [NIM] of Annaly is, of course, strongly correlated with the 20 yrs- 3 mos Treasury bond yield spread: r=0.825. The interest income of Annaly increases with the Treasury yield spread: this was fully expected, but needed to be verified because it was an underlying assumption here. Similarly, the correlation between dividend and same-quarter NIM is very high at r=0.915.
Next, I plotted the dividend yield as a function of the BV (again from the results of 44 quarters, ignoring my 1 outlier. All the numbers come from the 10Q forms on Annaly's website). There is an undeniable relationship between BV and dividend yield. Historically, a lower BV corresponds to a lower dividend yield.
The correlation coefficient stands at r=0.757. Again, it is dangerous to read too much from a simple correlation analysis (which ignores the specific reasons for any change in BV), but there is a stronger relationship between dividend yield and rate spread than between dividend yield and BV (0.875 vs. 0.757). In short, BV has had, historically, only a second-order impact on the dividend compared to the yield spread. The Treasury spread is a better predictor of the dividend than the BV. The correlations between BV and yields at different quarterly intervals are (using 45 data points for same quarter, 1, 2, and 3 quarters of difference, respectively): r=0.761, 0.736, 0.671, and 0.598. Like with the rate spread, the BV at any given quarter has an almost immediate impact on the dividend, and this impact vanishes after a few quarters.
What change in BV can we expect between the 1st and 2nd quarter of 2013 for NLY? Many investors and analysts much more competent than me in this field have estimated the BV at the end of June 2013. I will only quote two results: analysts at Wells Fargo Securities estimate the 2nd quarter BV to be $13.12, and fellow Seekingalpha contributor, David White, estimated on June 21 this BV to be $13.82. Compared to the reported BV of $15.19 on March 31, 2013, these two values give a 9% to 13.6 % quarter-to-quarter drop.
Over the 45 quarters studied, only 4 times the quarter-to-quarter BV dropped by more than 9%, resulting in an average dividend cut at the end of the quarter by 5.25 cents (an average over only 4 values may not be significant). By contrast, there were 11 quarters where the yield spread rose by more than 9% (relatively to the previous quarter), resulting in an average 0.45 cents raise in the dividend that same quarter, and followed by an average raise of 5.9 cents the next quarter and by 8.9 cents the quarter after. Historically, large drops in BV have been followed by immediate dividend cuts, while a significant widening in the yield spread is followed by a significant dividend raise only at the next quarter (and even more so the quarter after that). If the past is any guide to the present, this means that we can reasonably expect a dividend hike as early as the 3rd quarter of 2013, but even more so during the 4th quarter. Let's not forget that during the 2nd quarter of 2013, the relative raise in yield spread was a whopping 20% (from 2.66% to 3.18%).
Another possible relationship is between dividend yield and the leverage used by Annaly. You might expect a mREIT to increase its dividend when leverage increases: if it takes more risks, there should be more rewards. However, the correlation coefficient is r=-0.729, a rather strong anti-correlation: high leverage historically went hand-in-hand with low dividends. The correlation between leverage at a given quarter and dividend yield at the next quarter is r=-0.776, and only r=-0.753 for a two-quarter interval. It appears that the leverage is reflected in the dividend mostly after at least one quarter.
This strong dependence between leverage and yield is not too surprising as it reflects, amongst others, that the Annaly management team often used a higher leverage when the interest rate spread was narrow (to increase its interest income) and reduced the leverage when the spread widened. The following figure partly confirms this pattern:
There is also a strong anti-correlation between BV and leverage: r=-0.917. A high leverage corresponds to a low BV, which is what you expect from the definition of leverage. The BV per common share is related to the company's net asset value [NAV], i.e. the difference between assets and liabilities. Leverage, the debt-to-equity ratio, can be written as the liabilities divided by the NAV. The company can increase leverage by taking more debts, which immediately reduces its NAV and BV (until this debt is used to purchase new assets). Another way to increase leverage is for the value of the company's assets to decrease (while the liabilities stay put): then BV and NAV decrease and the leverage mechanically increases (this is what happened for mREITs these past few months).
Another parameter that is known to negatively impact the income of mREITs is the constant prepayment rate [CPR]. A quarterly CPR of X% means that during this quarter, X% of the mortgage pool principal was paid off. With a CPR of 48% for instance, the highest value ever reached by NLY (and our outlier of 3rd quarter of 2003), almost half of the mortgage assets were paid off during the quarter: the management team then had to reinvest this money by purchasing new MBS paying different interest rates (possibly lower ones). Thus, for a CPR of X% a mREIT has X% of the principal paid off but will never receive some of the interest money it was expecting. Not only a mREIT loses interest income due to prepayment, but if the MBS was bought at a premium in the first place it also means a loss in the fair value of this asset.
With the upcoming rise in long-term rates, the CPR is expected to decrease as there is less incentive for homeowners to refinance (traditionally one of the main source of prepayment). In NLY's case, there appears to have been historically little correlation between the dividend yield and the CPR during the same quarter, as is shown on the next figure:
(click to enlarge)
The correlation coefficient is very low, which is easily seen on the figure (the scatter in the data points is quite large). However, if instead of comparing dividend yield and CPR for the same quarter the correlation with the dividend yield taken a few quarters away is computed, the result is different:
Again, the error bars are large on each data point, and as the quarterly interval increases less data points are used: recent quarterly results are ignored. This creates a bias and makes the values of r for large quarterly intervals less reliable. Still, there is a general trend that shows that the CPR is anti-correlated with the dividend: high CPRs are linked to low dividends. Moreover, the maximum correlation (in absolute value) is obtained for a rather large time interval: the CPR at a given quarter does not immediately impact the dividend, but by forcing the mREIT to modify its portfolio it impacts the dividend several quarters down the road. As the CPR is expected to decrease in the upcoming quarters thanks to the rise in long-term rates, this should have a positive impact on the dividend.
Finally, it must be mentioned that in this section I looked at the dividend yields instead of the dividends themselves, but very similar correlation values are obtained with the dividends rather than their yields. Another point is that the four variables tested are not independent: BV, CPR, and leverage are all linked to the interest rates.
What do these correlations teach us?
I believe that this analysis, even though it is based on simple correlations, it has biases and is by no means perfect, quantitatively confirms what was qualitatively known by many investors: that the primary predictor of Annaly's dividends is the interest rate spread. A wide spread means high dividends. And it does not take long for the dividend to adjust to the new spread. The book value also impacts the dividend: low book values have historically been linked to low dividend yields. This was once again verified during the 2nd quarter of 2013, when Annaly cut its dividend by 5 cents following the rapid ascent of long-term interest rates and the consequent drop in its book value.
Logically then, if the BV stabilizes in the coming months (i.e. if long-term interest rates stabilize or change at a slower pace than this past quarter), then Annaly should be able to quickly raise its dividend by taking advantage of a wider yield spread. We can expect this hike as early as the 3rd quarter of 2013, but more likely in the 4th quarter. It is not the change in the interest rates that is detrimental to book value so much as it is the speed at which the rates change: too fast a change and the hedges are not as efficient at protecting the BV. The prospect of a rising dividend in upcoming quarters is further supported by the following fact:
Annaly's book value is (partly) protected
Indeed, the BV is partly protected through hedges and the level of protection keeps increasing. The notional amount of interest-rate swaps taken by Annaly covered 46% of its MBS portfolio at the end of March 2013, a historical high point, as can be seen on the following figure:
Before 2005, the 10Q forms do not report any swap amounts: I am not sure what to make of it, as I find it hard to believe that Annaly would not contract any swaps to hedge against changes in the rates. What is sure though, is that the management team has not been caught completely like a deer in headlights by the recent rise in rates: they anticipated this event (though probably not the speed at which it unfolded) a long time ago and have systematically increased the notional amount of their swaps since 2008. It is reasonable to assume that the trend continued this quarter. In fact, if we extrapolate the trend from recent quarters, it is likely that about 50% of the MBS value is now covered by swaps (which does not mean that the drop in book value will only be 50% of what it would have been without hedges). If long-term interest rates change at a slower pace in the coming months, rather than violently seesawing, this level of hedges should do a good job at protecting the book value. It costs money to hedge the portfolio, but it is worth it in times of rising rates.
Annaly is cheap right now
It really is. At a share price of $12.57 at closure of the normal trading hours on June 28, and assuming that the two BV estimates I previously quoted are correct (granted, it's a big if), that gives a price-to-book-value at the end of the 2nd quarter of 2013 in the range 0.909-0.958. By historical standards, this is cheap: I plotted the usual price-to-BV ratio as a function of time in the following figure (I only used 1 data point per quarter --- the value at the end of the quarter --- so the curve is sub-sampled. Still, it gives a reasonable idea of the historical variations).
The NLY share is cheap, and is trading below its likely current BV. But cheap does not mean undervalued: the share could accurately reflect the current value of Annaly and its future prospects. In my own opinion, for what it is worth, this is not the case. The prospects of Annaly are rather bright in the mid-term (until at least 2015 when the Fed may start raising short-term interest rates), provided that the long-term rates stabilize or climb slowly: the market reacted violently when the Federal Reserve hinted at a tapering of its quantitative easing program, but this reaction was very unusual by historical standards and therefore it is not unreasonable to expect that coming changes in interest rates will be more subdued.
A simple correlation analysis confirmed that the Treasury yield spread is the best predictor of Annaly's dividend, of the four parameters studied. Book value and CPR have a strong impact too (a delayed one for the CPR). If you add the prospects of rising dividends in the coming quarters, with a reasonable probability that it happens before the end of 2013 (again, provided the long-term rates change at a slower pace from now on), to the likely discount compared to current book value of the share, then the recent share price of NLY definitely makes for a good entry point.
Yes, Annaly is risky: if long-term interest rates continue to experience violent convulsions like they have these past 2 months, it could be disastrous for the BV. The leverage will only compound this issue. In my opinion though, there is more upside than downside currently associated with NLY for the patient investor who does not mind collecting dividends while waiting for the share price to rise again. Annaly has increased its level of interest rates swaps, has a leverage that is on the low end of its historical range, and the Fed guaranteed near-zero short-term interest rates until at least 2015.
A last word regarding the leverage of mREITS in general and Annaly in particular: at the end of 1st quarter, it was at 6.6 (defined as the debt-to-equity ratio). Since 2002, the leverage of NLY stayed in the range 5.4 to 11.5. Here is how it compares to the leverage of a few select banks: at the end of 1st quarter 2013, Barclays reported an adjusted gross leverage of 20, Wells Fargo reported a Tier 1 leverage of 9.53, and Bank of America a Tier 1 leverage of 7.56. In short, Annaly has a lower leverage than some well regarded financial institutions (which function in a roughly similar manner: banks borrow money on the short-term to lend at long-term rates. Banks arbitrage the difference in interest rates). Therefore, I think it is somewhat disingenuous to present the mREITs as inherently dangerous due to their high leverage, without emphasizing the fact that with respect to leverage alone they are not more dangerous than some better known financial institutions from which investors do not hesitate to buy shares: mREITs are risky, but many investors put money in other companies that are as, or more, risky using leverage as a risk metric.
For people who are, legitimately, worried by the leverage of Annaly, it is worth remembering that other mREITs have a lower leverage: TWO for instance only had a 3.1 debt-to-equity ratio on March 31, 2013, and its share price lost about 16% in 3 months, compared to the roughly 25% of NLY. There is also PMT, one of my personal favorites, with a leverage as low as 1.3 at the end of last quarter: PennyMac Mortgage Investment Trust is more than a mREIT, but it got equally punished recently just because (it seems) Mortgage Investment is attached to its name. These make for interesting plays too: the long-term upside might be less than for NLY, but the downside is probably more limited.