Entering text into the input field will update the search result below

NLY Timing System Update

Oct. 01, 2014 9:20 PM ETNLY3 Comments
Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

I have updated the regression equation for NLY, reported in my recent article "Timing the Purchase and Sale of Mortgage REITs, using the statistics from the end of the third quarter. The 10-day MA of TYX (30-year T-bond yield) was 3.267, the FFR (federal funds rate) was unchanged at .125, and the 10-day MA of NLY was 11.17. The close on 9/30 was $10.675.

The regression starting from the last quarter of 1997 yielded the result:

Pred. NLY = 12.98 - 1.32(FFR) + 1.054(TYX). The r-sq was .428, a drop from .448 the previous quarter, and the standard error rose from 2.728 to 2.783.

Therefore, the predicted value for MA10NLY is 16.26, and the current value of 11.17 is 1.83 standard errors lower than the predicted value. The stock is very cheap.

The current yield spread between the FFR and TYX is 3.14, and nowhere near the sell trigger of a yield spread above 4.50.

As you may recall I back-tested a system in which it bought NLY if the yield spread inverted, and it didn't sell until the spread was at least +1.0. In my back-test, I had bought at close the day after these signals triggered. I tried doing it on the same day that they triggered, and there was no significant difference.

Buy signals were also triggered if NLY's price fell more than one standard error below the value predicted by the regression (as long as the yield spread was not above 4.5), and a sell signal was triggered when the price went above 1.5 standard errors above the predicted value, unless there had been a yield inversion and the yield spread was still less than one. I executed the buy and sell signals at the close of the day AFTER NLY's price fell below YHAT minus 1 SE or above YHAT plus 1.5 SE. Now I have tried doing it the same day that the signal triggers.

For example, on 3/07/08, MA10TYX was 4.558, and FFR was 3.0. The predicted value for NLY was 16.80. The standard error for the prediction was 1.89, and so YHAT minus 1 standard error was 14.91. The next day (3/10/08) NLY closed at $14.58, but in my original back-test I waited another day and bought at the close at a price of $17.08. Why did I wait a day? Because the value of MA10TYX and FFR might have changed on 3/10/08, changing the predicted value of NLY, such that there was no buy. As it turns out that was silly.

If I buy at the close on the day that the preceding day's data says that NLY's price is less than YHAT minus 1 SE at close, and sell on the same day that it exceeds YHAT plus 1.5 SE, I was able to greatly improve the return of the system. Since the end of 2002, the CAGR shot up from 13.96% to 16.35%. The buy-and-hold with dividend re-investment was 6.58% since the last day of 2002. The improved return was primarily because of two buy signals. One signal was the one illustrated in the example when the system bought at $14.58 on 3/7/08 instead of at $17.08 on 3/10/08, and the other signal occurred on 8/8/11 buying at $16.34, instead of $17.86 a day later.

I also tried buying intra-day at exactly -1 SE and selling intra-day at +1.5 SE, but that didn't help the return at all. So buy or sell near the close if it looks like the close will be under the buy level or above the sell level.

Currently the sell price for NLY is at $20.43, so it looks like I'm a long-term holder, unless TYX jumps a lot, and triggers the yield spread sell signal of +4.5%. I view this as highly unlikely because the end of QE in the US and the start of QE in Europe is deflationary for the US, and I expect our T-bond yields to decline. Our yields are much higher than most of Europe because their inflation is not as high, but their bonds are not as safe as ours, and they are trying to bring on inflation. The falling Euro will provide a little inflation for them. My bond holdings made it a good day for me today. I also bought PTY during the Bill Gross panic. It has a 9% yield and its dividends don't fluctuate.

If we approach a buy or sell level I will inform you in the comments section to this Instablog while I'm alive and in possession of my faculties. I'm currently working on adding XLF and NLY's NAV to the regression equation, and I am also working on moving window regressions. The latter is where I drop the oldest observation from the data when I add the most recent one. So I start with a regression on the data from 12/31/97 to 12/31/02. Then the next regression would be for the data from 3/31/98 to 3/31/03 etc. I don't have great expectations for the results, but if anyone knows how I can combine =Linest with edit-fill-down in Excel, it would save me a lot of time. Write me in the comments section. I also have SPSS if anyone knows how I can do it there in one run, without asking for each individual regression.

Analyst's Disclosure: The author is long NLY, REM, TLT, PTY.

Seeking Alpha's Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body.

Recommended For You