Thanks for your reply! I started my backtest in July 2009 which is around the inception date of UPRO. I may try to look into simulated data of UPRO prior to 2009. I think we have reached a consensus that one should tilt towards UPRO in tax-advantaged accounts.skierincolorado wrote: ↑Thu Sep 30, 2021 12:54 pmWell first I'd make my usual point which I'm sure you know if you've read my other posts. I don't love HFEA because of the high fees and because LTT are very inferior to ITT. That being said, we're talking about the same thing in terms of keeping similar returns across our accounts. There's no way to make more money in the Roth without also taking more risk in the Roth. Of course over very long time horizons, the risk should pay off. So from that perspective, it would be best to do mostly TMF in taxable, and 100% UPRO in Roth.dc93 wrote: ↑Thu Sep 30, 2021 11:24 amHi skierincolorado,skierincolorado wrote: ↑Wed Sep 29, 2021 10:34 pmI'm not sure I see your point.adamhg wrote: ↑Wed Sep 29, 2021 9:57 pm
It doesn't work this way at the portfolio level unless you plan to backdoor Roth on your first rebalance. You need each separate, impermeable account type weighted independently. Otherwise when UPRO tanks, you're left without your bond portion in your 401k and you're effectively 3x levered in SPY and can't access the TMF sitting in your Roth to rebalance. Now try your back tests with each asset class independent and see if it matches your expectations
Let's say I have 100k TMF in IRA and 100k UPRO in taxable. If the UPRO tanked, I would sell TMF in the IRA and buy UPRO in the IRA. The same as if they were in the same account.
I do find it kind of an unneccesary risk in terms of what if stocks do much better than bonds the next 30 years, or vice-versa. My taxable account might do much better than my IRA. That's an avoidable risk. Personally I have a have 5 accounts (4 account types) and in most of them I hold stocks and bonds, but some of them are just stocks. So I make some effort to spread the risk around across accounts, but I don't worry about it too much and I definitely don't try to make it equal.
Thanks for your excellent post! I have a question that may be relevant here:
What's the most tax-efficient way to implement HFEA in two or more accounts?
I had an earlier post in the HFEA thread: viewtopic.php?p=6110148#p6110148
Here I assumed that initially we put an equal amount of money ($10k) in taxable and Roth IRA, both in a 55/45 split. When we rebalance, we never touch the taxable part and only adjust the portion in Roth to make the overall asset allocation 55/45.
Ideally, I want more money in Roth IRA if possible, and avoid large realized gains in the taxable account (that could happen when there is a huge imbalance between the two accounts that making roth 100% on one of UPRO/TMF wouldn't suffice for the purpose of rebalancing).
The backtest result of this strategy (back to 2009) suggest that it works fine for 12 years, and ~65% money does end up in Roth. But there are many moving pieces in my experiment that has the potential to improve the tax efficiency:
(1) What if we have more cash flow? (e.g. $6000 yearly contribution to Roth)
(2) What if the amounts in Roth/taxable are unequal?
(3) Will this strategy break down in a longer time horizon?
(4) Instead of 55/45 on both accounts, should we start with a different allocation?
I don't have a good intuition about the answers to these questions, maybe I should just do more back-testing. I would really appreciate comments and suggestions if folks here are interested.
Best,
dc93
I think your backtest would breakdown. For example, if we start with 10k in each and balanced 55/45 in both, it would only take a ~50% drop in UPRO before you were 100% UPRO in your Roth. If UPRO dropped 50% again, as it did in 2008-9, your Roth would now be quite small. Of course in the long-run it should pay off since UPRO should rise more. So I think this scenario is fine.
The worst scenario I think is if UPRO went up very quickly initially, you would end up 100% TMF in the Roth. In this case, I would probably rebalance in taxable to get some UPRO back into the Roth. To avoid this, I think it would be best to start with more of the UPRO in the roth. This is consistent with the initial point of taking more risk in the Roth.
When did your backtest start in 2009? I am surprised it avoided this scenario. I think if you started in March 2009 or April 2020, you'd end up 100% TMF in the Roth, or close to it, which would be undesirable long term.
I'd probably start 80/20 UPRO/TMF in the Roth, and 30/70 UPRO/TMF in taxable. Or 70/30 and 40/60. Basically tilting to UPRO in Roth to avoid this scenario.
Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Are you using "portfolio margin" on IBKR?skierincolorado wrote: ↑Wed Sep 29, 2021 10:15 pm Since it's a taxable account with margin, you don't need to hold cash collateral for futures. It just adds to your margin requirements. In your example with 550k equity, and holding 550k domestic + 230k intl + 400k ZF, if it's Reg T, the margin req would be .25*780k + ~1k per ZF. Since the 550k in equity is much more than that, you'd have a lot of excess liquidity. You'd have a 230k margin loan at 1.5-2.5% depending on which IBKR plan you have. This would drop to ~.7% if/when you write a 230k box spread.
I was reading IBKR docs and it have somewhat confusing phrase about futures collateral on portfolio margin configuration:
Margin is determined on a real-time basis with immediate position liquidation if the minimum margin requirement is not met. Only cash may be used to meet variation margin requirements. Special 50% of variation margin benefits are made available during the day on some futures exchanges. Non-US futures options are available to US legal resident customers. You must have futures trading permissions in order to have options on futures trading permissions.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
https://www.cmegroup.com/tools-informat ... ytics.html
/ZT = 0.08%
/ZF = 0.31%
/ZN = 0.64%
/ZB = 1.28%
/UB = 1.91%
They move around a LOT by the hour/day as prices change. That's about as high as I've seen them since I started casually monitoring a few weeks ago.
55% VUG - 20% VEA - 20% EDV - 5% BNDX
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
HLTCM wrote: ↑Thu Sep 30, 2021 4:16 pm
https://www.cmegroup.com/tools-informat ... ytics.html
/ZT = 0.08%
/ZF = 0.31%
/ZN = 0.64%
/ZB = 1.28%
/UB = 1.91%
They move around a LOT by the hour/day as prices change. That's about as high as I've seen them since I started casually monitoring a few weeks ago.
Last edited by hdas on Tue Oct 26, 2021 6:33 pm, edited 1 time in total.
....
- millennialmillions
- Posts: 187
- Joined: Sun Aug 18, 2019 6:03 pm
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Circling back on this to ensure I have a fundamental understanding. I'm seeing conflicting information on the right number to use as your measure of exposure when calculating leverage. Bentonkb mentions either invoice price or settlement price. In this example, Skier uses the par value. In another example at the bottom of this post, it appears a different value is being used.millennialmillions wrote: ↑Wed Sep 29, 2021 9:43 pm Am I thinking amount the amount of ZF correctly? I want $400,000 of exposure, so I buy 4 contracts with a par value of $100,000 each? Bentonkb mentions in this post that a lot of people use the settlement price as the notional value?
I am thinking the right number to use is the market value to get closest to the gains/losses you'd have if you actually held the amount in ITT. ZF market value is currently around $122,300. So if I want $400,000 in ITT exposure I'd buy 3 ZF and $33,000 worth of VGIT. Does that sound right?
In case it's not clear, I'm a statistician by training slowly trying to learn the world of finance.
-
- Posts: 17
- Joined: Fri Sep 03, 2021 6:18 pm
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
I believe you are correct. If you want to be precise ZF's duration is shorter than VGIT so you'll have to adjust for that.millennialmillions wrote: ↑Thu Sep 30, 2021 9:07 pmCircling back on this to ensure I have a fundamental understanding. I'm seeing conflicting information on the right number to use as your measure of exposure when calculating leverage. Bentonkb mentions either invoice price or settlement price. In this example, Skier uses the par value. In another example at the bottom of this post, it appears a different value is being used.millennialmillions wrote: ↑Wed Sep 29, 2021 9:43 pm Am I thinking amount the amount of ZF correctly? I want $400,000 of exposure, so I buy 4 contracts with a par value of $100,000 each? Bentonkb mentions in this post that a lot of people use the settlement price as the notional value?
I am thinking the right number to use is the market value to get closest to the gains/losses you'd have if you actually held the amount in ITT. ZF market value is currently around $122,300. So if I want $400,000 in ITT exposure I'd buy 3 ZF and $33,000 worth of VGIT. Does that sound right?
In case it's not clear, I'm a statistician by training slowly trying to learn the world of finance.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
sorry, what's funny? If I made a mistake just let me know. I was reporting the implied repo in the post. Its not negative at the moment but it was last week.hdas wrote: ↑Thu Sep 30, 2021 7:05 pmlol, this is not it dude,........but the info appears to be there, the Implied Repo, which is negative!!LTCM wrote: ↑Thu Sep 30, 2021 4:16 pmhttps://www.cmegroup.com/tools-informat ... ytics.html
/ZT = 0.08%
/ZF = 0.31%
/ZN = 0.64%
/ZB = 1.28%
/UB = 1.91%
They move around a LOT by the hour/day as prices change. That's about as high as I've seen them since I started casually monitoring a few weeks ago.
55% VUG - 20% VEA - 20% EDV - 5% BNDX
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
If we combine the idea of Lifecycle investing and Risk-parity, maybe we don't need much leverage on bond after all.
Assuming Bob has an income of 1M for 30 years and then retires. To borrow against future income, if he use futures to gain leverage, it's not risk free to him since the duration of his debt do not match the duration of his income. To adjust the duration of his debt, he could short some treasuries. Maybe short 100% 15-year treasuries.
If he uses a 30/70 stock/ITT strategy and leverages it by 300%, he would long 210% ITT. The long ITT part and the short 15-year treasuries part mostly cancel out. Therefore, he long 90% stock and call it a day.
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Why would you want to duration match your debt risk-free? The point of investing and the reason it generates return, is to take risk. Just like we should take risk in equities, we should also take interest rate risk in the bond market. The return of bonds is *always* a delta between risk-free rates on cash and rates on longer durations. This is true whether we are 70 years old and putting up our own cash, or 30 years old and borrowing cash. The term for this is carry.daze wrote: ↑Fri Oct 01, 2021 5:48 amIf we combine the idea of Lifecycle investing and Risk-parity, maybe we don't need much leverage on bond after all.
Assuming Bob has an income of 1M for 30 years and then retires. To borrow against future income, if he use futures to gain leverage, it's not risk free to him since the duration of his debt do not match the duration of his income. To adjust the duration of his debt, he could short some treasuries. Maybe short 100% 15-year treasuries.
If he uses a 30/70 stock/ITT strategy and leverages it by 300%, he would long 210% ITT. The long ITT part and the short 15-year treasuries part mostly cancel out. Therefore, he long 90% stock and call it a day.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
My understanding is that lifecycle investing is about capitalizing future incomes. Ideally, you borrowed every pennies of future incomes into present value and repay them with all future incomes. The risk-free status is the baseline, then you decided what risk you would take.skierincolorado wrote: ↑Fri Oct 01, 2021 10:05 amWhy would you want to duration match your debt risk-free? The point of investing and the reason it generates return, is to take risk. Just like we should take risk in equities, we should also take interest rate risk in the bond market. The return of bonds is *always* a delta between risk-free rates on cash and rates on longer durations. This is true whether we are 70 years old and putting up our own cash, or 30 years old and borrowing cash. The term for this is carry.daze wrote: ↑Fri Oct 01, 2021 5:48 amIf we combine the idea of Lifecycle investing and Risk-parity, maybe we don't need much leverage on bond after all.
Assuming Bob has an income of 1M for 30 years and then retires. To borrow against future income, if he use futures to gain leverage, it's not risk free to him since the duration of his debt do not match the duration of his income. To adjust the duration of his debt, he could short some treasuries. Maybe short 100% 15-year treasuries.
If he uses a 30/70 stock/ITT strategy and leverages it by 300%, he would long 210% ITT. The long ITT part and the short 15-year treasuries part mostly cancel out. Therefore, he long 90% stock and call it a day.
Assuming a hypothetical lender grants 23M dollars to Bob today, in exchange for all his 30 years salaries. The debt is guaranteed to be repaid no matter what. Now Bob could invest the 23M dollars at will. He could put it in a traditional 60/40 portfolio, or he put it in 90/210 risk-parity portfolio.
In real life, Bob probably could not get such a deal. ( Well.. Maybe Bob was hit by a car and loses his future working ability, and he gets a 23M dollars settlement.) But we could try to mimic it, and cancel out what could be cancelled out to reduce the implementation costs.
The question is whether a (20.7M stock/-20.7M cash) portfolio is similar enough to a (20.7M stock/48.3M ITT/-23M 30-year debt ladder/-46M cash) portfolio, if we consider the reduced cost?
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.millennialmillions wrote: ↑Thu Sep 30, 2021 9:07 pmCircling back on this to ensure I have a fundamental understanding. I'm seeing conflicting information on the right number to use as your measure of exposure when calculating leverage. Bentonkb mentions either invoice price or settlement price. In this example, Skier uses the par value. In another example at the bottom of this post, it appears a different value is being used.millennialmillions wrote: ↑Wed Sep 29, 2021 9:43 pm Am I thinking amount the amount of ZF correctly? I want $400,000 of exposure, so I buy 4 contracts with a par value of $100,000 each? Bentonkb mentions in this post that a lot of people use the settlement price as the notional value?
I am thinking the right number to use is the market value to get closest to the gains/losses you'd have if you actually held the amount in ITT. ZF market value is currently around $122,300. So if I want $400,000 in ITT exposure I'd buy 3 ZF and $33,000 worth of VGIT. Does that sound right?
In case it's not clear, I'm a statistician by training slowly trying to learn the world of finance.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Thanks for sharing the twitter post. It shows that coupon yield contributes the most to bond return. ITT does have more roll yield than LTT but the contribution to the total return is tiny. Does this mean we should still use LTT? I'd like to hear Skier's thoughts on this.Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 amI don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.millennialmillions wrote: ↑Thu Sep 30, 2021 9:07 pmCircling back on this to ensure I have a fundamental understanding. I'm seeing conflicting information on the right number to use as your measure of exposure when calculating leverage. Bentonkb mentions either invoice price or settlement price. In this example, Skier uses the par value. In another example at the bottom of this post, it appears a different value is being used.millennialmillions wrote: ↑Wed Sep 29, 2021 9:43 pm Am I thinking amount the amount of ZF correctly? I want $400,000 of exposure, so I buy 4 contracts with a par value of $100,000 each? Bentonkb mentions in this post that a lot of people use the settlement price as the notional value?
I am thinking the right number to use is the market value to get closest to the gains/losses you'd have if you actually held the amount in ITT. ZF market value is currently around $122,300. So if I want $400,000 in ITT exposure I'd buy 3 ZF and $33,000 worth of VGIT. Does that sound right?
In case it's not clear, I'm a statistician by training slowly trying to learn the world of finance.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
- millennialmillions
- Posts: 187
- Joined: Sun Aug 18, 2019 6:03 pm
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Thanks for the explanation. Are you sure that the Treasury Conversion Factor is used to get to a "market" price that would be paid at delivery? Reading the CME Group paper on TCF, "A conversion factor is the approximate decimal price at which $1 par of a security would trade if it had a six percent yield-to-maturity."Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
In my (uninformed) reading of this, it seems to be going the opposite direction: the TCF is used to adjust the price to what it would be if it had a 6% yield to maturity. This is useful to compare the various deliverable securities apples-to-apples. I could definitely be wrong; would be great to get your thoughts or hear what number others are using here.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Good twitter thread. I think the approximation is to pick the steepest part of the yield curve to maximize the carry. It is around 5 - 7 years now. Maybe if you consider cost of borrowing it is /ZN is the best with 6.5 years of CTD (and 6.3 duration)Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am
I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Hi, Klaus. I need a little bit of help to understand the twitter thread. You are right that ITT has the best roll yield, but it also shows that roll yield only contributes a little to total return. It is the coupon yield that contributes the most. They also said that coupon yield is the most predictive of future return. Is my interpretation wrong? Thanks.klaus14 wrote: ↑Wed Oct 06, 2021 5:57 amGood twitter thread. I think the approximation is to pick the steepest part of the yield curve to maximize the carry. It is around 5 - 7 years now. Maybe if you consider cost of borrowing it is /ZN is the best with 6.5 years of CTD (and 6.3 duration)Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am
I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
"E.g., the conversion factor for delivery of the 2-3/8%millennialmillions wrote: ↑Tue Oct 05, 2021 8:44 pmThanks for the explanation. Are you sure that the Treasury Conversion Factor is used to get to a "market" price that would be paid at delivery? Reading the CME Group paper on TCF, "A conversion factor is the approximate decimal price at which $1 par of a security would trade if it had a six percent yield-to-maturity."Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
In my (uninformed) reading of this, it seems to be going the opposite direction: the TCF is used to adjust the price to what it would be if it had a 6% yield to maturity. This is useful to compare the various deliverable securities apples-to-apples. I could definitely be wrong; would be great to get your thoughts or hear what number others are using here.
T-note of Aug-24 vs. December 2017 10-year T-note
futures is 0.8072. This suggests that a 2-3/8% security
is approximately valued at 81% as much as a 6%
security. Assuming a futures price of 125-08+/32nds (or
125.265625 expressed in decimal format), the principal
invoice amount may be calculated as follows.
Principal Invoice Price
= 125.265625 x 0.8072 x $1,000
= $101,114.41
E.g., the conversion factor for delivery of the 1-7/8%
T-note of Aug-24 vs. December 10-year T-note
futures is 0.7807. This suggests that a 1-7/8% security
is approximately valued at 78% as much as a 6%
security. Assuming a futures price of 125-08+/32nds
(or 125.265625), the principal invoice amount may be
calculated as follows.
Principal Invoice Price
= 125.265625 x 0.7807 x $1,000
= $97,794.87
In order to arrive at the total invoice amount, one must of
course further add any accrued interest since the last semi
annual interest payment date to the principal
invoice amount."
Quoted from https://www.cmegroup.com/education/file ... utures.pdf
The TCF is used both ways. You take the cash price of the bond and divide by TCF to get the futures settlement price. The calculation to go the other way is demonstrated above.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
I think you are right. Historically, roll yield was small compared to coupon. But nowadays coupon is also small so around 7 years roll yield is almost same as coupon.cometqq wrote: ↑Wed Oct 06, 2021 8:56 amHi, Klaus. I need a little bit of help to understand the twitter thread. You are right that ITT has the best roll yield, but it also shows that roll yield only contributes a little to total return. It is the coupon yield that contributes the most. They also said that coupon yield is the most predictive of future return. Is my interpretation wrong? Thanks.klaus14 wrote: ↑Wed Oct 06, 2021 5:57 amGood twitter thread. I think the approximation is to pick the steepest part of the yield curve to maximize the carry. It is around 5 - 7 years now. Maybe if you consider cost of borrowing it is /ZN is the best with 6.5 years of CTD (and 6.3 duration)Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am
I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Thanks, Klaus. But in the long run, LTT might be a better choice. This is probably why HFEA uses LTT, but in the short-term, holding LTT will be very painful...klaus14 wrote: ↑Wed Oct 06, 2021 9:36 amI think you are right. Historically, roll yield was small compared to coupon. But nowadays coupon is also small so around 7 years roll yield is almost same as coupon.cometqq wrote: ↑Wed Oct 06, 2021 8:56 amHi, Klaus. I need a little bit of help to understand the twitter thread. You are right that ITT has the best roll yield, but it also shows that roll yield only contributes a little to total return. It is the coupon yield that contributes the most. They also said that coupon yield is the most predictive of future return. Is my interpretation wrong? Thanks.klaus14 wrote: ↑Wed Oct 06, 2021 5:57 amGood twitter thread. I think the approximation is to pick the steepest part of the yield curve to maximize the carry. It is around 5 - 7 years now. Maybe if you consider cost of borrowing it is /ZN is the best with 6.5 years of CTD (and 6.3 duration)Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am
I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Are you talking historically or currently? The graph in the twitter shows roll is about half the return of the 7 year currently. Historically, it would be less because the coupon was higher. However, after you factor in borrowing costs, roll would still be about half the historical return. The return of a an asset should alwasy be measured relative to the risk free return. This is this risk premium. The risk premium, both now and historically, has come about half from roll. LTT have very little extra return, both now and historically, for the immense extra risk - which is why they are a poor investment. Owning more ITT has always beaten owning an equivalent amount of risk in LTT over any time horizon of more than a few years.cometqq wrote: ↑Wed Oct 06, 2021 8:56 amHi, Klaus. I need a little bit of help to understand the twitter thread. You are right that ITT has the best roll yield, but it also shows that roll yield only contributes a little to total return. It is the coupon yield that contributes the most. They also said that coupon yield is the most predictive of future return. Is my interpretation wrong? Thanks.klaus14 wrote: ↑Wed Oct 06, 2021 5:57 amGood twitter thread. I think the approximation is to pick the steepest part of the yield curve to maximize the carry. It is around 5 - 7 years now. Maybe if you consider cost of borrowing it is /ZN is the best with 6.5 years of CTD (and 6.3 duration)Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am
I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
The only reason HFEA uses LTT is because you can get sufficient duration by owning an ETF (TMF) which makes it easier. Historically, owning the same total duration of ITT outperforms HFEA by a mile. It's not even close. You can own significantly less total duration, and take less risk, and still have higher long-term returns than HFEA. The backtests in this thread show this to be true over many historical periods, such as 1955-present, 1978-present and 1991-present. And none of these backtests have even considered the significant cost savings that owning futures have relative to TMF.cometqq wrote: ↑Wed Oct 06, 2021 11:03 amThanks, Klaus. But in the long run, LTT might be a better choice. This is probably why HFEA uses LTT, but in the short-term, holding LTT will be very painful...klaus14 wrote: ↑Wed Oct 06, 2021 9:36 amI think you are right. Historically, roll yield was small compared to coupon. But nowadays coupon is also small so around 7 years roll yield is almost same as coupon.cometqq wrote: ↑Wed Oct 06, 2021 8:56 amHi, Klaus. I need a little bit of help to understand the twitter thread. You are right that ITT has the best roll yield, but it also shows that roll yield only contributes a little to total return. It is the coupon yield that contributes the most. They also said that coupon yield is the most predictive of future return. Is my interpretation wrong? Thanks.klaus14 wrote: ↑Wed Oct 06, 2021 5:57 amGood twitter thread. I think the approximation is to pick the steepest part of the yield curve to maximize the carry. It is around 5 - 7 years now. Maybe if you consider cost of borrowing it is /ZN is the best with 6.5 years of CTD (and 6.3 duration)Bentonkb wrote: ↑Tue Oct 05, 2021 7:19 am
I don't think that $122,300 is the correct value to use in your calculations. The settlement price of $122,300 is the hypothetical price of the cheapest-to-deliver bond at an (arbitrary) yield of 6%. The reference yield was chosen back when typical bond yields were higher. There is some discussion of lowering the reference yield as they have done in the past. It is purely a matter of convenience to allow each bond in the deliverable basket to be compared on a level playing field. It has nothing to do with how much you would have to pay if you were long on the contract and took delivery at expiration.
The invoice price is the amount you would have to pay at delivery. It is the futures quote * $1000 * TCF. The TCF is the fudge factor that converts between the actual yield and the reference yield of 6%. The /ZF TCF is 0.7999 today, so the risk equivalent of a /ZF contract is $122,300 * 0.7999, which is the market price for a $100,000 par bond with a 0.5% coupon, 0.9% yield, and a maturity of 2/28/2026. (I ran this through the PV calculation in Excel. It is pretty close.)
Using the par value is only off by 2%, so it is better than using the settlement price of $122,300.
Please help me figure out if I'm doing this wrong because the correction factors are bigger for the /ZB contract that I'm using this quarter. The par value for /ZB is still $100k, but the invoice is about $147k, and the settlement is about $160k. Those are significant correction factors, IMO.
I know Skier will object to the LTT. I am researching the idea of switching to ITT instead. Here is an interesting Twitter thread on the topic:
https://twitter.com/choffstein/status/1 ... wsrc%5Etfw
Owning 20 years of duration of LTT currently has a return of 2% including roll. Owning 20 years of duration of ZN has a yield of 6% including roll. Same duration risk. 3x the return.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Thanks, skier. I've spent quite a bit of time reading this thread. Lots of good stuff! I've recently replaced all my TMF with TYA (it is unfortunate that my account is restricted from trading futures). We will see how this goes.skierincolorado wrote: ↑Wed Oct 06, 2021 11:12 amThe only reason HFEA uses LTT is because you can get sufficient duration by owning an ETF (TMF) which makes it easier. Historically, owning the same total duration of ITT outperforms HFEA by a mile. It's not even close. You can own significantly less total duration, and take less risk, and still have higher long-term returns than HFEA. The backtests in this thread show this to be true over many historical periods, such as 1955-present, 1978-present and 1991-present. And none of these backtests have even considered the significant cost savings that owning futures have relative to TMF.cometqq wrote: ↑Wed Oct 06, 2021 11:03 amThanks, Klaus. But in the long run, LTT might be a better choice. This is probably why HFEA uses LTT, but in the short-term, holding LTT will be very painful...klaus14 wrote: ↑Wed Oct 06, 2021 9:36 amI think you are right. Historically, roll yield was small compared to coupon. But nowadays coupon is also small so around 7 years roll yield is almost same as coupon.cometqq wrote: ↑Wed Oct 06, 2021 8:56 amHi, Klaus. I need a little bit of help to understand the twitter thread. You are right that ITT has the best roll yield, but it also shows that roll yield only contributes a little to total return. It is the coupon yield that contributes the most. They also said that coupon yield is the most predictive of future return. Is my interpretation wrong? Thanks.
Owning 20 years of duration of LTT currently has a return of 2% including roll. Owning 20 years of duration of ZN has a yield of 6% including roll. Same duration risk. 3x the return.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hhhhh
Last edited by hdas on Tue Oct 26, 2021 6:34 pm, edited 1 time in total.
....
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Like you, I'm guessing those are quarterly. Where are you getting them from?hdas wrote: ↑Wed Oct 06, 2021 11:38 amHow do you arrive at this calculations?. The current roll yield (Dec-March) is:skierincolorado wrote: ↑Wed Oct 06, 2021 11:12 am
Owning 20 years of duration of LTT currently has a return of 2% including roll. Owning 20 years of duration of ZN has a yield of 6% including roll. Same duration risk. 3x the return.
ZF 0.5% 6:1
ZN 0.64% 4:1
UB 0.47% 1:1
- Since this is per rolling period, I'm assuming you multiply * 4 to get an estimate, with the assumption that next roll is similar?
- I'm assuming you also multiply for the # contracts according to the hedge ratio correct?
In any case, using the current roll yield numbers, I can't see a way to arrive at your numbers. H
I was using the graph of yield in the twitter thread which shows ~2% for both 7 year and 20 year duration.
It would take 3 7-yr bonds to have the same total duration risk as a 20-yr bond. So 2%*3 = 6%.
Using your numbers, assuming they are accurate and quarterly, it would be ~2%(.47*4) for UB and ~7.5%(.64*4*3) for an equivalent amount of ZN. I think UB has the duration of about 3 ZN, not 4.
I don't know what date the twitter graph is from... that could explain the difference
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Do you have the ZT roll yield too please?
55% VUG - 20% VEA - 20% EDV - 5% BNDX
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Yeah I saw TYA in the HFEA thread. It fixes the fee problem with TMF and TYD, but you still have the same lack of leverage that TYD has. For modestly leveraged strategies, TYA is probably fine though. I'm only modestly leveraged, so I actually could use TYA and UPRO. I'm basically 45% UPRO / 55% TYA.
Last edited by skierincolorado on Wed Oct 06, 2021 12:12 pm, edited 1 time in total.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hskierincolorado wrote: ↑Wed Oct 06, 2021 12:00 pm
Like you, I'm guessing those are quarterly. Where are you getting them from?
I was using the graph of yield in the twitter thread which shows ~2% for both 7 year and 20 year duration.
It would take 3 7-yr bonds to have the same total duration risk as a 20-yr bond. So 2%*3 = 6%.
Using your numbers, assuming they are accurate and quarterly, it would be ~2%(.47*4) for UB and ~7.5%(.64*4*3) for an equivalent amount of ZN.
I don't know what date the twitter graph is from... that could explain the difference
Last edited by hdas on Tue Oct 26, 2021 6:34 pm, edited 1 time in total.
....
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
I have worked out roll yield myself multiple times in this thread you will see. I also just did it using the current 7 year and 5 year rates. I get .93% roll yield + 1.3% annual coupon = 2.23% total return for 7 year duration.hdas wrote: ↑Wed Oct 06, 2021 12:11 pmIt's not quarterly, the roll "yield" reflects the accumulated %.skierincolorado wrote: ↑Wed Oct 06, 2021 12:00 pm
Like you, I'm guessing those are quarterly. Where are you getting them from?
I was using the graph of yield in the twitter thread which shows ~2% for both 7 year and 20 year duration.
It would take 3 7-yr bonds to have the same total duration risk as a 20-yr bond. So 2%*3 = 6%.
Using your numbers, assuming they are accurate and quarterly, it would be ~2%(.47*4) for UB and ~7.5%(.64*4*3) for an equivalent amount of ZN.
I don't know what date the twitter graph is from... that could explain the difference
I get those numbers from the settlement of the futures, as of yesterday
Perhaps nobody should take assumptions for granted from a twitter thread if they haven't worked out the numbers themselves.
Read the paper below, which makes good sense, and let's work out HOW does an investor crystalizes this "yield". IMO is very dubious, and people should focus on the exposure they want and what they do with the free cash. That's mostly what matters.
Also, it's clear that the movement of the curve is very very crucial for this strategy to work, and historically is all back related to yields, when curve flattens, the additional risk of LTT is not worth taking vis a vis the yield of the short maturity and AT THAT POINT you are better off being in the short end.
Financial Analysts Journal
Derivatives
The “Roll Yield” Myth
Hendrik Bessembinder
Pages 41-53 | Published online: 12 Dec 2018
Download citation https://doi.org/10.2469/faj.v74.n2.5 CrossMark LogoCrossMark
Full Article Figures & data References Citations Metrics Reprints & Permissions Get access
Abstract
Futures investors are frequently said to periodically pay or receive the difference in futures prices across contracts with different delivery dates. But this “roll yield” is mythical: No such cash flow occurs—at the time of roll trades or on any other date. However, although the term is a misnomer, the roll yield does contain useful information. It explains when futures gains exceed or fall short of spot-price changes, and for storable assets, it provides information regarding benefits to the marginal holder of a spot position. This article clarifies the actual role of the roll yield.
We've also discussed a lot how changes in the yield curve mean these returns are unlikely to be realized. These calculations assume no change in the yield curve. In reality, interest rate forward contracts indicate the market expects rates to increase. We've also done calculations of expected return given market expected future interest rates. Expected returns on both intermediate and long-term bonds are substantially less than even their coupons would indicate. Probably under .5%. I would be pleasantly surprised if ITT or LTT yielded more than 1% over the next year or two.
Calculating current roll yield, expected future returns, trying to time the yield curve, etc... these are all interesting to me but ultimately pointless and unlikely to be actionable. Both backtests and theory show diversifying across stocks and bonds to have superior risk-adjusted returns, and for short and intermediate duration bonds to have superior risk-adjusted returns to longer durations. This investment strategy has proven to be near-optimal across many decades. Diversifying between stocks and bonds is solidly supported by modern portfolio theory, and preferring shorter duration bonds to longer durations is solidly supported by bet against beta theory.
Nobody in the HFEA thread calculates roll for TMF. And a few try to time either the bond or stock market, or both - but are quickly and rightfully shot down. The one mistake HFEA makes is that LTT (TMF) have poor risk-adjusted returns both historically and theoretically (bet against beta). This is likely because you can't get enough duration from bonds in a 3x ETF other than LTT. We can get the same negative correlation with stocks, but higher risk-adjusted returns, using ITT futures.
These discussions are interesting. But ultimately we should look at the big picture and not try to time the market. The big picture, based on both theory and backtesting, is 1) diversify across stocks and bonds and 2) ITT > LTT
Last edited by skierincolorado on Wed Oct 06, 2021 12:35 pm, edited 2 times in total.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hskierincolorado wrote: ↑Wed Oct 06, 2021 12:21 pm Both backtests and theory show diversifying across stocks and bonds to have superior risk-adjusted returns, and for short and intermediate duration bonds to have superior risk-adjusted returns to longer durations.
Last edited by hdas on Tue Oct 26, 2021 6:35 pm, edited 1 time in total.
....
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
There are a number of papers posted in this thread that show the returns of investing in futures are the same as investing in funds. Minus fees of course, which is a point in favor of futures.
The optimal ratio of stocks to bonds is always changing and can never be known in advance. This has been well discussed in this thread. The best thing to do is pick a ratio, and overall amount of leverage, which has been durable historically. I really don't care whether I get the precise correct ratio. The return of owning stocks and bonds has always been vastly superior, whether you have the exact correct ratio or not.
Or perhaps you are referring to rebalancing frequency? The vastly suprior returns of stocks+ITT are durable across rebalancing frequencies.
All of the backtests in this thread use dynamic fianancing costs.
The collateral requirements of futures contracts are trivial and earn cash rates in broker account and nothing in an IRA. Rates are near zero anyways. Even at 1% interest rates, 100k in an IRA in a 130/200 stock/ITT portfolio would have a cash drag of a mere $150. If rates rise substantially above 2%, then yes I would look for a broker that allows holding the cash collateral in STT.
85k VTI
15k cash as collateral for:
45k MES
200k ZF
Last edited by skierincolorado on Wed Oct 06, 2021 12:55 pm, edited 3 times in total.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Do you understand the constantly fluctuating IRR rates on futures contracts? Does it have anything to do with the financing cost of buying futures? Is it a good indicator of the financing cost or is it merely an indicator of the supply/demand for certain securities in a futures contract and which is the CTD?
If not relevant to financing cost then how does one discover the implied financing costs of different treasury futures contracts?
I don’t understand it and would appreciate some insight.
55% VUG - 20% VEA - 20% EDV - 5% BNDX
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hhhhh
Last edited by hdas on Tue Oct 26, 2021 6:35 pm, edited 1 time in total.
....
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
This is likely because the borrowing cost is substantially above 3m libor since 2012. For a fully funded position there would be a .3% drag on CAGR relative to a fund, assuming .05% fee for the fund and borrowing at .35% over 3m LIBOR, holding collateral in 3m LIBOR. Since even in a highly leveraged position at least half the stock exposure is through a fund, the actual drag on the portfolio would be half of that. So yes, you can factor in a .15% drag on CAGR in any backtest. These differences are trivial. Are you proposing an alternative? The only alternative I see is to not leverage which would be foolish given the difference in CAGR is 10%+ in any long-term backtest. Why quibble over .15%?hdas wrote: ↑Wed Oct 06, 2021 1:00 pm
Kindly me show me which paper you are talking about. (not cme or pimco pamphlets, since these are interested parties). For instance, I've looked deeply at the difference between holding ES futures vs a fund. And the difference is substantial and can't be recovered by the investment of the collateral, in the case of the fully funded position.
This ad from Invesco for their S&P500 ETF finds the costs of futures were .47% over the last 5 years. But this is an ad for their ETF, and they use a number of pessimistic assumptions. They use the lower overnight rate on cash instead of 3m Libor for the collateral. It's not clear whether they pick appropriate times to roll ("1-9 weeks before expiry"). https://etf.invesco.com/sites/default/f ... 18_3.0.pdf
I assume you're aware that treasury futures have much lower financing costs. You can see from figure 5 here that the financing cost averages under .1% and is often negative. https://www.financialresearch.gov/brief ... Trades.pdf
The fact that the exact ratio of ZF to UB to match interest rate risk is constantly changing (slightly) doesn't negate the validity of the backtest. The backtest shows ITT to be vastly suprior to LTT no matter what ratio we pick - with lower drawdowns and higher returns. In fact, we can (and *should* IMO) pick a ratio of ZF to UB that is lower than the current or historical ratios of their DV01 or the ratio of their durations. This would be consistently less risk. And yet the return is consistently higher.
For example, the ratios of their durations would suggest a 3:1 ratio. Their DV01s would suggest an even higher ratio. And yet we can easily use a much lower ratio - such as 2:1 (replacing LTT with ITT at a 2:1 ratio) and stocks/ITT will outperform stocks/LTT over any backtest over a long period. This is because of their vastly superior risk adjusted returns, consistent with beta against beta theory.
I said all the backtests use dynamic financing costs. I didn't say that they used perfectly accurate financing costs. Dynamic is not the same as perfectly accurate. The financing costs are dynamic and accurate within a couple tenths of a percent (within a tenth of a percent for treasuries which is where the large majority of the borrowing occurs). Some backtests simulated slightly higher financing rates. The conclusion that stock + ITT are superior to stocks only or stocks + LTT remains unchanged, durable across assumptions of financing costs, time period tested, and by a large margin.hdas wrote: ↑Wed Oct 06, 2021 1:00 pmNo they are not.....sometimes financing is above 3m libor, sometimes is below.skierincolorado wrote: ↑Wed Oct 06, 2021 12:47 pm All of the backtests in this thread use dynamic fianancing costs.
I don't believe that pursuing alpha is worthwhile and the capital efficiency of using futures is superior to the alternatives.hdas wrote: ↑Wed Oct 06, 2021 1:00 pmI guess I should have been more precise. I meant the excess cash that a futures position "frees up". Aka Portable Alpha, capital efficiency, etc.skierincolorado wrote: ↑Wed Oct 06, 2021 12:47 pm The collateral requirements of futures contracts are trivial and earn cash rates in broker account and nothing in an IRA. Rates are near zero anyways. Even at 1% interest rates, 100k in an IRA in a 130/200 stock/ITT portfolio would have a cash drag of a mere $200.
80k VTI
20k cash as collateral for:
70k MES
200k ZF
- millennialmillions
- Posts: 187
- Joined: Sun Aug 18, 2019 6:03 pm
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Thank you for the reference; this really cleared it up for me. The underlying asset of the contract can be assumed to be the CTD security. The true market value of this security is the invoice price. It seems like you'd really want the Total Invoice Amount rather than the Principal Invoice Price, but they should be close enough to ignore the accrued interest.Bentonkb wrote: ↑Wed Oct 06, 2021 9:04 am "E.g., the conversion factor for delivery of the 2-3/8%
T-note of Aug-24 vs. December 2017 10-year T-note
futures is 0.8072. This suggests that a 2-3/8% security
is approximately valued at 81% as much as a 6%
security. Assuming a futures price of 125-08+/32nds (or
125.265625 expressed in decimal format), the principal
invoice amount may be calculated as follows.
Principal Invoice Price
= 125.265625 x 0.8072 x $1,000
= $101,114.41
E.g., the conversion factor for delivery of the 1-7/8%
T-note of Aug-24 vs. December 10-year T-note
futures is 0.7807. This suggests that a 1-7/8% security
is approximately valued at 78% as much as a 6%
security. Assuming a futures price of 125-08+/32nds
(or 125.265625), the principal invoice amount may be
calculated as follows.
Principal Invoice Price
= 125.265625 x 0.7807 x $1,000
= $97,794.87
In order to arrive at the total invoice amount, one must of
course further add any accrued interest since the last semi
annual interest payment date to the principal
invoice amount."
Quoted from https://www.cmegroup.com/education/file ... utures.pdf
The TCF is used both ways. You take the cash price of the bond and divide by TCF to get the futures settlement price. The calculation to go the other way is demonstrated above.
Based on this, my actual exposure to ITT through 4 ZF is about $392,000 when I thought it was $489,000. Conveniently, this is almost exactly the difference of one ZF...time to buy another one!
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
What is the drawbacks of using SPY futures nominated in foreign currency (non-USD)?
Is implied financing rate would be different in this case?
Is implied financing rate would be different in this case?
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
intuitively shouldn't be. a foreign bank can hold SP500 stocks and just quote the value in EUR. Their financing cost is still USD Libor.
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
i remember confirming that it is the latter ($489,000) but i don't remember enough to argue heremillennialmillions wrote: ↑Wed Oct 06, 2021 5:46 pmThank you for the reference; this really cleared it up for me. The underlying asset of the contract can be assumed to be the CTD security. The true market value of this security is the invoice price. It seems like you'd really want the Total Invoice Amount rather than the Principal Invoice Price, but they should be close enough to ignore the accrued interest.Bentonkb wrote: ↑Wed Oct 06, 2021 9:04 am "E.g., the conversion factor for delivery of the 2-3/8%
T-note of Aug-24 vs. December 2017 10-year T-note
futures is 0.8072. This suggests that a 2-3/8% security
is approximately valued at 81% as much as a 6%
security. Assuming a futures price of 125-08+/32nds (or
125.265625 expressed in decimal format), the principal
invoice amount may be calculated as follows.
Principal Invoice Price
= 125.265625 x 0.8072 x $1,000
= $101,114.41
E.g., the conversion factor for delivery of the 1-7/8%
T-note of Aug-24 vs. December 10-year T-note
futures is 0.7807. This suggests that a 1-7/8% security
is approximately valued at 78% as much as a 6%
security. Assuming a futures price of 125-08+/32nds
(or 125.265625), the principal invoice amount may be
calculated as follows.
Principal Invoice Price
= 125.265625 x 0.7807 x $1,000
= $97,794.87
In order to arrive at the total invoice amount, one must of
course further add any accrued interest since the last semi
annual interest payment date to the principal
invoice amount."
Quoted from https://www.cmegroup.com/education/file ... utures.pdf
The TCF is used both ways. You take the cash price of the bond and divide by TCF to get the futures settlement price. The calculation to go the other way is demonstrated above.
Based on this, my actual exposure to ITT through 4 ZF is about $392,000 when I thought it was $489,000. Conveniently, this is almost exactly the difference of one ZF...time to buy another one!
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
-
- Posts: 17
- Joined: Fri Sep 03, 2021 6:18 pm
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
CME Says It May Introduce 20-Year Treasury Futures Contracts (October 12, 2021) https://www.bloomberg.com/news/articles ... s-contract
CME Group Inc., the dominant provider of futures tied to U.S. Treasuries, took a step toward introducing contracts tied to the 20-year bonds that were brought back in 2020 by the federal government.
The Chicago-based exchange has designed three prototypes and will unveil them in a webinar on Oct. 19, according to a notice on its website announcing the event. Following that, CME will “determine the validity and design preferences of this potential new product.”
The U.S. resumed selling 20-year Treasuries in May 2020 after a more than three-decade hiatus.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hskierincolorado wrote: ↑Wed Oct 06, 2021 1:51 pm doesn't negate the validity of the backtest. The backtest shows ITT to be vastly suprior to LTT no matter what ratio we pick - with lower drawdowns
Last edited by hdas on Tue Oct 26, 2021 6:36 pm, edited 1 time in total.
....
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
What is the data source? There are a lot of complexities in these methodologies. Was the position fully funded or not?hdas wrote: ↑Thu Oct 14, 2021 2:48 pmI got my hands a 10y futures dataset going back to 2000. You seem to be already convinced of your idea, but perhaps other fellow travelers can exercise a bit of caution.skierincolorado wrote: ↑Wed Oct 06, 2021 1:51 pm doesn't negate the validity of the backtest. The backtest shows ITT to be vastly suprior to LTT no matter what ratio we pick - with lower drawdowns
I use the historical CME hedge ratio since 2009, and 3:1 for data before that.
I'm also pretty sure that you have leveraged the blue line incorrectly. The green line does not look simply like a leveraged verson of the blue line. Could you explain your methodology please? What financing cost did you use? It should be LIBOR - this does not imply that ZN financing costs are LIBOR. The blue line (if it is fully funded which it doesn't appear to be but shold be) is created by taking 100k and buying 100k exposure to ZN. The 100k is held in cash, earning LIBOR. But if we want to 2x leverage it, the next 100k doesn't earn LIBOR. Only the first 100k earned LIBOR. We simulate this by using LIBOR as the finance cost. This does NOT imply that ZN borrows at LIBOR. ZN may borrow at more than LIBOR, and that would be factored into the blue line already.
There's at least one mistake. The return of ZN since 2011 is 26.2%, while the return of VFITX is 25.7%.** Your graph shows ZN dramatically falling behind VFITX during this period. Most likely whatever data source you chose is not simulating a fully funded position, which it should be. The goal here is to compare a fully funded position in VUSTX to a partially funded 2x position in ZN. To do this, we need to start with a fully funded ZN position, and then leverage it 2x with the leverage incurring a borrowing cost on the half that is not funded.
I'm also nearl certain you have used some unrealistically high borrowing cost to leverage the blue line to the green line. Somehow little spikes are magnified nearly 10x from blue to green.
**Methodology is publicly available. 10 year return for ZN is from SPGlobal. 10 year return for VFITX is from PortfolioVisualizer.
https://www.spglobal.com/spdji/en/indic ... /#overview
https://www.portfoliovisualizer.com/bac ... ion1_1=100
There are two mistakes being made. The blue line is not fully funded. And you have used some incorrect financing cost to leverage blue to green.
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Here is ZN vs VFITX. Nearly identical.
Sources for data are SPGlobal ZN total return index and Portfolio Visualizer for VFITX. I graphed the returns side by side in excel.
Sources for data are SPGlobal ZN total return index and Portfolio Visualizer for VFITX. I graphed the returns side by side in excel.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hhhhh
Last edited by hdas on Tue Oct 26, 2021 6:36 pm, edited 2 times in total.
....
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Would you mind using a fixed 2x leverage on the blue line? Maybe using the 3-4x leverage is creating volatility decay?hdas wrote: ↑Thu Oct 14, 2021 5:55 pmThis data is from a friend, collected from CQG. Daily futures data, rolled quarterly, however, there could be a mistake in the rolling of the contracts. This is not complex, but I can't vouch for the quality of the data.skierincolorado wrote: ↑Thu Oct 14, 2021 3:12 pm What is the data source? There are a lot of complexities in these methodologies. Was the position fully funded or not?
The blue line is simply the price change of the futures + 90% of the fully funded position invested in a 1-2 year bond fund from Vanguard (VFISX), which is much better than 3month bills.skierincolorado wrote: ↑Thu Oct 14, 2021 3:12 pm I'm also pretty sure that you have leveraged the blue line incorrectly.
You don't need any financing assumptions because these are the historical prices of the future contract, all the implied financing is embedded in the price. For leverage modeling, I used daily returns x the hedge ratio of ZN vs UB that traditionally is 4:1 and 3:1skierincolorado wrote: ↑Thu Oct 14, 2021 3:12 pm The green line does not look simply like a leveraged verson of the blue line. Could you explain your methodology please? What financing cost did you use? It should be LIBOR - this does not imply that ZN financing costs are LIBOR.
The only thing I could have a mistake is that the data is corrupted. I have high quality data for UB futures. I can compare that with VUSTX.skierincolorado wrote: ↑Thu Oct 14, 2021 3:12 pm There's at least one mistake. The return of ZN since 2011 is 26.2%, while the return of VFITX is 25.7%.**
I do not know how much money are you putting into this
I still find the blue line to be suspect data. SPGlobal matches VFITX nearly identically and their methodology invests the fully funded position in LIBOR not 1-2 year T-Bills, which would be less favorable.
Are you saying that the blue line you are creating by taking your friends data, and then investing the fully funded amount in T-Bills, or does your friend's data do that already? Are you able to send me your friend's data? If you're taking the friend's data, and then simulating holding the fully funded amount in T-Bills, is it possible you are making a mistake in that process? How do you determine the fully funded amount? It's just that your result is not matching multiple public data sets for the return of ZN.
I have a significant amount in ITT futures. I'm still confident a mistake is being made somewhere but this is definitely of interest to me. As we saw, even that mutual fund that was advertising their fund to be better than investing in futures, they found that the return of futures was pretty close to their fund.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hskierincolorado wrote: ↑Thu Oct 14, 2021 6:12 pm I still find the blue line to be suspect data. SPGlobal matches VFITX nearly identically and their methodology invests the fully funded position in LIBOR not 1-2 year T-Bills, which would be less favorable.
Last edited by hdas on Tue Oct 26, 2021 6:37 pm, edited 1 time in total.
....
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hhhhh
Last edited by hdas on Tue Oct 26, 2021 6:37 pm, edited 1 time in total.
....
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hhhhh
Last edited by hdas on Tue Oct 26, 2021 6:37 pm, edited 1 time in total.
....
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Interesting. You said the UB futures data was a different source? The duration is different yeah, UB is longer. With investing of the funding it would probably look very similar to a slightly leveraged version of VUSTX.hdas wrote: ↑Thu Oct 14, 2021 6:23 pmI find that puzzling too, and it's likely that the data didn't adjust for rolling properly.skierincolorado wrote: ↑Thu Oct 14, 2021 6:12 pm I still find the blue line to be suspect data. SPGlobal matches VFITX nearly identically and their methodology invests the fully funded position in LIBOR not 1-2 year T-Bills, which would be less favorable.
Here's the UB futures vs VUSTX, without adding any funding to the position. I'm pleasantly surprised that the futures outperform the fund without investing the funding of the position!!!. I think it has to do with the differences of duration between the CTD's and the fund.
I don't know if this is the case for ZN.
-
- Posts: 2377
- Joined: Sat Mar 21, 2020 10:56 am
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Yeah your friend's data is definitely different than the SPGlobal data. There's also the WSJ data that comes from FactSet which is different from both of them. Will try to figure out the differences. I read through the SPGlobal methodology and it looked sound to me. I think I might have seen one or two other sources at some point too. I agree it would be good to settle on a reliable source and know the differences in methodology that are producing the differences. Tried to PM you my email but it looks like you have PM turned off?hdas wrote: ↑Thu Oct 14, 2021 6:32 pmHere's 2xskierincolorado wrote: ↑Thu Oct 14, 2021 6:12 pm
Would you mind using a fixed 2x leverage on the blue line? Maybe using the 3-4x leverage is creating volatility decay?
I still find the blue line to be suspect data. SPGlobal matches VFITX nearly identically and their methodology invests the fully funded position in LIBOR not 1-2 year T-Bills, which would be less favorable.
Are you saying that the blue line you are creating by taking your friends data, and then investing the fully funded amount in T-Bills, or does your friend's data do that already? Are you able to send me your friend's data? If you're taking the friend's data, and then simulating holding the fully funded amount in T-Bills, is it possible you are making a mistake in that process? How do you determine the fully funded amount? It's just that your result is not matching multiple public data sets for the return of ZN.
I added the returns of the 2 year vanguard fund, to simulate the fully funded position. I did not make a mistake in this, and my hunch is that the data is not properly roll adjusted. The fully funded amount I use is 90% of the notional, I don't think the mistake is there, those details are trivial.skierincolorado wrote: ↑Thu Oct 14, 2021 6:12 pm
Are you saying that the blue line you are creating by taking your friends data, and then investing the fully funded amount in T-Bills, or does your friend's data do that already?
Give me an email address and I can send you the data, maybe you find the mistake, if any.
H
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Has anyone done any analysis of this on an after-tax basis? I'm not really interested in leveraging my equities so all of my leverage would be in treasury futures (something like 90/10 or 80/20 stocks/futures collateral). My concern is getting marked-market every year makes holding futures in taxable much worse compared to equities in a taxable account, to the point where it might not even be worth the effort. Of course, this is not so bad at the moment with yields so low and not much expected tax impact, but I'm not too interested in changing my strategy based on the absolute level of yields.
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
hhhhh
Last edited by hdas on Tue Oct 26, 2021 6:38 pm, edited 1 time in total.
....
- millennialmillions
- Posts: 187
- Joined: Sun Aug 18, 2019 6:03 pm
Re: Modified versions of HFEA with ITT and Futures / Lifecycle Investing with Modern Portfolio Theory
Based on the posts from skier and the backtest results (assuming the above discussion on ZF gets resolved), I’m now convinced that I should hold ITT instead of LTT. I’ve also decided to go with an allocation of 140% equity (98% domestic, 42% international) and 140% ITT. I transferred my Roth IRA to IBKR so I can now hold futures. Here are my planned holdings:
A few remaining questions I would appreciate any input on:
A few remaining questions I would appreciate any input on:
- Thoughts on my allocation across accounts? I don’t have any treasuries in my “Other Rtmt” accounts because they can’t trade futures, and it seems any other mechanism of treasury exposure is subpar.
- The participants of this thread have still not reached consensus on the correct amount to use when determining treasury allocation. Based on the CME resource Bentonkb provided, I believe it should be the Principal Invoice Price (quoted futures price * the conversion factor), so that’s what I’m using. MES seems more straightforward – just 5 * the index value, right?
- How does everyone decide the buffer on your futures margin requirement you’re comfortable with? E.g. if I have 14 MES contracts in a Roth IRA and I want to be able to withstand a 30% drop, I need 14 * 4,400 * 5 * 0.3 = $93,000 extra cash beyond the initial margin requirement. But what about the possibility of changing margin requirements, since they increase with market volatility?
- How does everyone hold their “buffer” cash that is not currently used to meet margin requirements in retirement accounts? Do you put it in a fund like ICSH, JPST, or MINT? Is there a way to ensure that would be liquidated first in case extra cash is needed to meet margin requirements?