2014 Trading Thread
12-08-2014
, 11:28 PM
12-09-2014
, 12:41 AM
Hey guys, I have a theory that I had been thinking about today in terms of long time horizon investing in tax advantaged accounts that don't allow the use of margin. The idea is to utilize leverage in the form of purchasing SPY LEAPS to determine if the decay/other factors are worthwhile given the benefits.
I broke it down into 2 hypothetical $100k portfolios, each consisting solely of positions in SPY for equities and DLTNX (I always talk about this fund and the manager David Gundlach, and its easy to use here) for bonds. I used 10 year data (though 8% is SPY's historical return anyway) on SPY to estimate an EV of an 8% annual return with a standard deviation of 14.66 and sharpe ratio of 0.49. Since DLTNX has only been around for a few years, I took Gundlach's previous fund that he built and ran, TGMNX, and used a much lower expected return of 5.5% as I am considerably less optimistic about fixed income returns in the future and didn't think using the higher historical returns was reasonable, and used an averaged DLTNX/TGMNX standard deviation and sharpe for 3 and 1.5 respectively.
Portfolio A has an 80/20 exposure to SPY/DLTNX. For its weighted totals calculations, I adjusted for a 100% total using 80% weighting to the SPY position and 20% weighting to the DLTNX position.
Portfolio B is long the equivalent of $100k worth of SPY -- or pretty close as I wanted to choose a slightly lower delta to account for gamma accelerating on the more historically likely scenario of SPY going up. The delta is reasonable to being long about 484 shares of SPY with $100k. For this I used 5 December 2016 $130 calls at a mid-price of $76.75, which utilizes $38,375 in cash. The plan would be to buy them 2 years out, and roll them after 1 year to keep theta down. The expected return was modeled on the basis of what happens to calls' value if SPY increases 8% in 1 year. Since there is inherently a time premium in holding options long, it ends up losing a bit more than it gains if SPY is down the same amount after 1 year. Obviously this also means that it will have gained less than it could have all else being equal. I took the average of these 2 figures (~19.1% on a gain, ~21.4% on a loss) and subtracted the difference from the gain (1.15%) to calculate the expected return. The expected return, 17.95%, is the ROC of the amount invested into the options, which represents the 8% return on the underlying SPY + the time premium and other variables in an option pricing model. For its weighted totals I adjusted for a 161.6% total - long 100% equity, long 61.6% bond fund, and thus used a weight of 0.62 and 0.38 respectively for the purposes of the new risk and return calculations.
It's worth noting that the standard deviation and sharpe ratio are irrelevant (but worth looking at to get an idea of the benefits), even though they are based on historical statistics. So long as both portfolios use the same variables, the end relative result should find the same relative improvement on a risk adjusted return basis.
Note that the results showed a higher expected return, lower standard deviation, and higher sharpe ratio. Let me know what you all think.
Here's a quick spreadsheet on it to make it easier to understand visually:
I broke it down into 2 hypothetical $100k portfolios, each consisting solely of positions in SPY for equities and DLTNX (I always talk about this fund and the manager David Gundlach, and its easy to use here) for bonds. I used 10 year data (though 8% is SPY's historical return anyway) on SPY to estimate an EV of an 8% annual return with a standard deviation of 14.66 and sharpe ratio of 0.49. Since DLTNX has only been around for a few years, I took Gundlach's previous fund that he built and ran, TGMNX, and used a much lower expected return of 5.5% as I am considerably less optimistic about fixed income returns in the future and didn't think using the higher historical returns was reasonable, and used an averaged DLTNX/TGMNX standard deviation and sharpe for 3 and 1.5 respectively.
Portfolio A has an 80/20 exposure to SPY/DLTNX. For its weighted totals calculations, I adjusted for a 100% total using 80% weighting to the SPY position and 20% weighting to the DLTNX position.
Portfolio B is long the equivalent of $100k worth of SPY -- or pretty close as I wanted to choose a slightly lower delta to account for gamma accelerating on the more historically likely scenario of SPY going up. The delta is reasonable to being long about 484 shares of SPY with $100k. For this I used 5 December 2016 $130 calls at a mid-price of $76.75, which utilizes $38,375 in cash. The plan would be to buy them 2 years out, and roll them after 1 year to keep theta down. The expected return was modeled on the basis of what happens to calls' value if SPY increases 8% in 1 year. Since there is inherently a time premium in holding options long, it ends up losing a bit more than it gains if SPY is down the same amount after 1 year. Obviously this also means that it will have gained less than it could have all else being equal. I took the average of these 2 figures (~19.1% on a gain, ~21.4% on a loss) and subtracted the difference from the gain (1.15%) to calculate the expected return. The expected return, 17.95%, is the ROC of the amount invested into the options, which represents the 8% return on the underlying SPY + the time premium and other variables in an option pricing model. For its weighted totals I adjusted for a 161.6% total - long 100% equity, long 61.6% bond fund, and thus used a weight of 0.62 and 0.38 respectively for the purposes of the new risk and return calculations.
It's worth noting that the standard deviation and sharpe ratio are irrelevant (but worth looking at to get an idea of the benefits), even though they are based on historical statistics. So long as both portfolios use the same variables, the end relative result should find the same relative improvement on a risk adjusted return basis.
Note that the results showed a higher expected return, lower standard deviation, and higher sharpe ratio. Let me know what you all think.
Here's a quick spreadsheet on it to make it easier to understand visually:
Last edited by DickFuld; 12-09-2014 at 12:49 AM.
