HEDGEFUNDIE's excellent adventure [risk parity strategy using 3x leveraged ETFs]

[staythecourse]

Last thing, we are waiting on the 1960-1980 data to see

What I really want to see is this analysis done on the data set from 1950-1970. That is more representative of what LIKELY will happen with interest rates going forward, i.e. starting low and slowly going up each and every year. Does ANYONE have the data for this period of time?

I love this concept, but just don't trust a data set where we know in RETROSPECT sp500 and LTT shot the lights out. Don't know where I read it, but think it was Dr. Bernstein in "TIAA" where he said the best diversifier to stocks from 1950-1970 was cash and not LTT. Makes sense as slowly increasing interest rates should have lower correlation with LTT. I am pretty sure I am not imagining that.

Either way I do love the idea. I still have not gotten an answer for stewardship issues if we have a 2008 scenario and the fund has HUGE outflows and gets shut down. Also, I love this idea if we were talking about it when the SP500 was NOT shooting out the lights.

Good luck.

Using Simba's backtesting spreadsheet, I get the following data for the unleveraged portfolio of 40/60 S&P500 and LTT for the time periods of 1950-1970, 1960-1980, and 1987-2018:

Code: Select all

Initial Investment				10000  10000   10000
Starting Year for backtest			1950	1960	1987
Ending Year for backtest			1970	1980	2018
# of Years backtesting				21	21	32
Offset 						79	89	116
Cycles Duration					30	30	30
					
MAR Benchmark Average	nominal			3.10%	5.40%	3.20%
					
WARNING: all metrics are derived from annual returns, therefore NOT taking in account intra-year events					
Time Period: 		varies			1950-70	'60-80	'87-18
Portfolio CAGR		nominal			6.11%	5.07%	8.86%
Portfolio CAGR		real			3.50%	-0.18%	6.10%
Portfolio Avg Return	nominal			6.30%	5.34%	9.21%
Portfolio Alpha		nominal			-4.46%	-2.70%	0.92%
					
Standard Deviation	nominal			6.62%	7.72%	8.90%
Stdev (Portfolio - MAR)	nominal			7.63%	8.34%	8.73%
Downside SD vs. MAR	nominal			3.03%	3.42%	1.37%
Upside SD vs. MAR	nominal			4.75%	3.20%	4.84%
Max. Drawdown		nominal			7.26%	12.69%	3.75%
Ulcer Index		nominal			1.62%	3.51%	0.98%
Maximum Withdrawal Rate	real			7.58%	5.59%	7.14%
Perpetual Withdrawal Rate real			3.90%	-0.21%	6.07%
					
Sharpe Ratio		nominal			0.42	-0.01	0.69
Sortino Ratio		nominal			1.06	-0.03	4.4
Ulcer Perf. Ratio 	nominal			1.98	-0.03	6.13
Portfolio Beta		nominal			1.22	1.2	0.86
					
US Mkt. Correlation	nominal			0.82	0.83	0.64
Int'l Mkt. Correlation	nominal			N/A	N/A	0.25
					
Total - Rebalanced	nominal			34727	28276	151495
Total - Unbalanced	nominal			57156	29450	137878
Total - Rebalanced	real			20592	9633	66456

Some thoughts:

1. I adjusted the time period from 1986 to 1987 in the '87-18 portfolio in column 3, because the TMF fund data provided by EfficientInvestor starts at June 1986 according to PV. To equalize everything, I started it in January 1987 here in Simba's spreadsheet. The CAGR of 8.86 here matches with a 40/60 VFINX (500) and VUSTX (LTT) portfolio in PV for that time period. https://www.portfoliovisualizer.com/bac ... tion2_1=60

2. Comparing the CAGRs of the 40/60 portfolio over those three time periods vs. 100% sp500 and 100% LTT:

Code: Select all

		'50-70	'60-80	'87-18
40/60		6.11%	5.07%	8.86%
100% S&P 500 	12.76%	7.70%	9.87%	
100% LTT	0.98%	2.69%	7.26%

3. Now the question becomes can we take the '50-70 unleveraged data and '60-80 unleveraged data above, and somehow match it to the leveraged data we have from '87-18 via some sort of ratio, even if somewhat imprecise? For example, we know that an unleveraged 40/60 portfolio from '87-18 returned 8.86% CAGR. Had it been leveraged, it would have returned 23.52%. So, can we say that during this time, the leveraged portfolio returned 2.65 times as much? (23.52/8.86=2.65)

If so, can we then extrapolate (or some other word, it's early in the morning), that 2.65 multiplier for the other time periods? So, for the '50-70 data, the unleveraged CAGR of 6.11% would have been 16.22% had it been leveraged? Or the '60-80 unleveraged CAGR of 5.07%--might it have been 13.46% had it been leveraged? Probably not. But someone should be able to come up with an estimate that takes into account such things as the difference in standard deviation, etc. in the data above.

Note that older data for LTT is given only in annual terms, and PV apparently requires at a minimum, monthly returns to come up with a benchmark it'll recognize.

Thanks for the work. I think the concept of this idea is great, but has mentioned to have a fair analysis more data is needed. I am very tech handicapped so really am hoping folks out there will do the number crunching to come up with the LEVERAGED returns.

Good luck.

[unclescrooge]

Last thing, we are waiting on the 1960-1980 data to see

What I really want to see is this analysis done on the data set from 1950-1970. That is more representative of what LIKELY will happen with interest rates going forward, i.e. starting low and slowly going up each and every year. Does ANYONE have the data for this period of time?

I love this concept, but just don't trust a data set where we know in RETROSPECT sp500 and LTT shot the lights out. Don't know where I read it, but think it was Dr. Bernstein in "TIAA" where he said the best diversifier to stocks from 1950-1970 was cash and not LTT. Makes sense as slowly increasing interest rates should have lower correlation with LTT. I am pretty sure I am not imagining that.

Either way I do love the idea. I still have not gotten an answer for stewardship issues if we have a 2008 scenario and the fund has HUGE outflows and gets shut down. Also, I love this idea if we were talking about it when the SP500 was NOT shooting out the lights.

Good luck.

Using Simba's backtesting spreadsheet, I get the following data for the unleveraged portfolio of 40/60 S&P500 and LTT for the time periods of 1950-1970, 1960-1980, and 1987-2018:


WARNING: all metrics are derived from annual returns, therefore NOT taking in account intra-year events

To simulate the returns of 3x leveraged fund you would need daily returns. It's the sideways choppiness that can erode the value. Annual returns are less useful.

[tmcc]

in a sideways / up tilt market , this strat will do well.

if there is a major crash, you will not recover vs the underlying index for a long time.

look at XBI vs LABU since May 2015. XBI is the SPDR biotech index. LABU is the 3x long ETF that tracks the same biotech index. in late 2015, the index got crushed. Using may 2015 as a starting point, LABU is still down -70%... XBI is up 3.5%.

as a leverage lover, i say best of luck. id do 2x SPY , VUG, SCHX and pay margin interest long before id touch another 3x fund.

[HomerJ]

The entire premise of my strategy rests upon the idea that when one asset collapses, the other asset rises sharply. If you don’t believe this will happen reliably going forward, you should not undertake this strategy. I am comfortable on this point, both because of historical returns and the underlying risk story of why this occurs.

Again, you are only looking at a period of time where long-term bonds were basically in a bull market 90% of the time with falling interest rates.

You've seen that UPRO can lose a ton of money and take 17 years to recover. And you expect the long-term treasury returns to make up for that. But I doubt they will make up for any UPRO losses as STRONGLY starting from low interest rates like today.

I mean, your strategy may work, but it is very unlikely to mimic the past 30 years. Your backtest was dependent on conditions that do not exist today.

Again, today's conditions aren't bad. The strategy could work. But it will NOT work in the same way. Your back-test has very little predictive power about what kind of returns you will get going forward.

[staythecourse]

The entire premise of my strategy rests upon the idea that when one asset collapses, the other asset rises sharply. If you don’t believe this will happen reliably going forward, you should not undertake this strategy. I am comfortable on this point, both because of historical returns and the underlying risk story of why this occurs.

Again, you are only looking at a period of time where long-term bonds were basically in a bull market 90% of the time with falling interest rates.

You've seen that UPRO can lose a ton of money and take 17 years to recover. And you expect the long-term treasury returns to make up for that. But I doubt they will make up for any UPRO losses as STRONGLY starting from low interest rates like today.

I mean, your strategy may work, but it is very unlikely to mimic the past 30 years. Your backtest was dependent on conditions that do not exist today.

Again, today's conditions aren't bad. The strategy could work. But it will NOT work in the same way. Your back-test has very little predictive power about what kind of returns you will get going forward.

Agreed. This seems to be the consensus. We need more data then a time period where we already know LT Treasury and Sp500 did very well. Sounds like a broken record by replicating a similar strategy starting after WWII and slowly up to 1970 or so would be good proxy (if any exists) of what is likely to happen next 30 years. A falling interest rate environment is not the same as a slowly rising one in its effect on price of a long term bond.

Good luck.

[siamond]

Historically, it can be shown that by holding uncorrelated pair of assets (e.g. S&P 500 & long Treasuries) and levering up, one could achieve a similar volatility profile with greater returns than by holding the S&P alone.

Are you assuming a negative correlation (when one zigs, the other tends to zag) or a null correlation (when one zigs, the other keeps a mind of its own)? Those are VERY different things.

[DosCommas]

Just how inefficient would this strategy be in a taxable account? What would one have to do to calculate if it would be worth doing based on a backtest? I am considering this might be an interesting use for play money but holding these funds in taxable and rebalancing them might offset the theoretical additional return, no? 35% federal 10.3% state...

[siamond]

The UPRO data is based on daily returns of VFINX and includes a 1% per year expense ratio (I subtract .01/250 for each day). I initially used ^GSPC as the data source, but forgot that I had switched to using VFINX because it accounts for the dividend. If you only use data from ^GSPC, your proxy will underperform UPRO. For each day, I did a weighted average of the close value (20%) and adjusted close value (80%). I determined the weighting using an iterative process until I found something that closely matched the actual UPRO performance.

Are you sure UPRO uses the S&P 500 total return index and not the price index? I could not find a clear answer on ProShares site. It sort of sounds like you are torturing the data until it confesses.

The best I can tell is that it’s somewhere in the middle. UPRO obtains a broad exposure to the index by putting a large portion of funds into individual stocks or unleveraged ETFs. Then, with a small portion of the funds, they use futures and swap contracts to hit the 3x daily goal. By holding the individual stocks and ETFs, they generate a dividend which is currently around 0.5% per Yahoo. Due to this dividend, UPRO is achieving better performance than just 3x the daily ^GSPC (or daily close of VFINX) performance. However, since they can’t apply leverage to that dividend, you can’t assume they will meet the 3x daily performance of the adjusted close of VFINX. Therefore, the actual performance of UPRO is somewhere in the middle. Enter the use of the weighted average. The 20/80 blend that I am using matches actual performance of UPRO quite well over the last 9+ years, so I feel comfortable using that blend going further in the past. If you want to be more conservative, feel free to just use the ^GSPC data without adjusting for dividend.

Although I can appreciate the explanation, it does seem like force fitting to me. There might be all sorts of other frictions happening which makes the UPRO returns significantly less than the "3x daily" goal (incl. transaction costs). I don't have a better idea though. What do you do with TMF? Same problem, right?

[siamond]

Agreed. This seems to be the consensus. We need more data then a time period where we already know LT Treasury and Sp500 did very well. Sounds like a broken record by replicating a similar strategy starting after WWII and slowly up to 1970 or so would be good proxy (if any exists) of what is likely to happen next 30 years. A falling interest rate environment is not the same as a slowly rising one in its effect on price of a long term bond.

Yes, totally agreed. I provided daily data series to EfficientInvestor (by PM), starting in Jan-73 for both the S&P 500 and Barclays LT Treasuries indices. I would much prefer to start in the early 60s, but this would require to use simulated data sets (e.g. the bond fund simulator we use for Simba) which would have at best monthly granularity.

One step at a time, let's explore the 1973+ data set first (I ran a coarse simulation myself and although the oil crisis was clearly NOT friendly to such 3x strategy, it doesn't seem to fall apart either -- but I don't master the whole 3x topic enough, so I'd rather wait for EfficientInvestor to run his own numbers). This whole thing IS an intriguing idea, in all fairness, worth exploring in more depth.

[HEDGEFUNDIE]

The entire premise of my strategy rests upon the idea that when one asset collapses, the other asset rises sharply. If you don’t believe this will happen reliably going forward, you should not undertake this strategy. I am comfortable on this point, both because of historical returns and the underlying risk story of why this occurs.

Again, you are only looking at a period of time where long-term bonds were basically in a bull market 90% of the time with falling interest rates.

You've seen that UPRO can lose a ton of money and take 17 years to recover. And you expect the long-term treasury returns to make up for that. But I doubt they will make up for any UPRO losses as STRONGLY starting from low interest rates like today.

I mean, your strategy may work, but it is very unlikely to mimic the past 30 years. Your backtest was dependent on conditions that do not exist today.

Again, today's conditions aren't bad. The strategy could work. But it will NOT work in the same way. Your back-test has very little predictive power about what kind of returns you will get going forward.

You are missing my point completely. I make no prediction about whether interest rates will go up or down in the long term, nor how much they will go up or down. Anyone who does should be ignored as a market timer, in the same way as those who predict future stock returns.

Let me state it one more time, as clearly as I can: my strategy rests on the daily correlation between stocks and long term bonds, NOT their medium-to-long term returns. As I am typing this right now, UPRO is down -0.56%, and TMF is up +1.31%. This is the behavior my strategy rests upon, and the behavior I do expect to continue, for risk and market structure reasons. It has nothing to do with long term interest rates or long term stock returns.

[rmelvey]

The entire premise of my strategy rests upon the idea that when one asset collapses, the other asset rises sharply. If you don’t believe this will happen reliably going forward, you should not undertake this strategy. I am comfortable on this point, both because of historical returns and the underlying risk story of why this occurs.

Again, you are only looking at a period of time where long-term bonds were basically in a bull market 90% of the time with falling interest rates.

You've seen that UPRO can lose a ton of money and take 17 years to recover. And you expect the long-term treasury returns to make up for that. But I doubt they will make up for any UPRO losses as STRONGLY starting from low interest rates like today.

I mean, your strategy may work, but it is very unlikely to mimic the past 30 years. Your backtest was dependent on conditions that do not exist today.

Again, today's conditions aren't bad. The strategy could work. But it will NOT work in the same way. Your back-test has very little predictive power about what kind of returns you will get going forward.

You are missing my point completely. I make no prediction about whether interest rates will go up or down in the long term, nor how much they will go up or down. Anyone who does should be ignored as a market timer, in the same way as those who predict future stock returns.

Let me state it one more time, as clearly as I can: my strategy rests on the daily correlation between stocks and long term bonds, NOT their medium-to-long term returns. As I am typing this right now, UPRO is down -0.56%, and TMF is up +1.31%. This is the behavior my strategy rests upon, and the behavior I do expect to continue, for risk and market structure reasons. It has nothing to do with long term interest rates or long term stock returns.

Have you thought about including gold in the portfolio? Or some other asset with low correlation to both stocks and bonds? The achilles heel of a stock/long duration bond portfolio is unexpected inflation, because it causes both stocks and bonds to go down in value.

[HEDGEFUNDIE]

Just how inefficient would this strategy be in a taxable account? What would one have to do to calculate if it would be worth doing based on a backtest? I am considering this might be an interesting use for play money but holding these funds in taxable and rebalancing them might offset the theoretical additional return, no? 35% federal 10.3% state...

Do the math, how much in taxes would you have to pay while growing $100k to $10M over 20 years?

[EfficientInvestor]

The UPRO data is based on daily returns of VFINX and includes a 1% per year expense ratio (I subtract .01/250 for each day). I initially used ^GSPC as the data source, but forgot that I had switched to using VFINX because it accounts for the dividend. If you only use data from ^GSPC, your proxy will underperform UPRO. For each day, I did a weighted average of the close value (20%) and adjusted close value (80%). I determined the weighting using an iterative process until I found something that closely matched the actual UPRO performance.

Are you sure UPRO uses the S&P 500 total return index and not the price index? I could not find a clear answer on ProShares site. It sort of sounds like you are torturing the data until it confesses.

The best I can tell is that it’s somewhere in the middle. UPRO obtains a broad exposure to the index by putting a large portion of funds into individual stocks or unleveraged ETFs. Then, with a small portion of the funds, they use futures and swap contracts to hit the 3x daily goal. By holding the individual stocks and ETFs, they generate a dividend which is currently around 0.5% per Yahoo. Due to this dividend, UPRO is achieving better performance than just 3x the daily ^GSPC (or daily close of VFINX) performance. However, since they can’t apply leverage to that dividend, you can’t assume they will meet the 3x daily performance of the adjusted close of VFINX. Therefore, the actual performance of UPRO is somewhere in the middle. Enter the use of the weighted average. The 20/80 blend that I am using matches actual performance of UPRO quite well over the last 9+ years, so I feel comfortable using that blend going further in the past. If you want to be more conservative, feel free to just use the ^GSPC data without adjusting for dividend.

Although I can appreciate the explanation, it does seem like force fitting to me. There might be all sorts of other frictions happening which makes the UPRO returns significantly less than the "3x daily" goal (incl. transaction costs). I don't have a better idea though. What do you do with TMF? Same problem, right?

Correct. The best methodology I can think of for TMF is to develop a weighted average of the close and adjusted close values of VUSTX. I don't claim it to be perfect. But for me, personally, it is close enough that it would be adequate enough to project what the performance of TMF would have been prior to its inception.

[HEDGEFUNDIE]

The entire premise of my strategy rests upon the idea that when one asset collapses, the other asset rises sharply. If you don’t believe this will happen reliably going forward, you should not undertake this strategy. I am comfortable on this point, both because of historical returns and the underlying risk story of why this occurs.

Again, you are only looking at a period of time where long-term bonds were basically in a bull market 90% of the time with falling interest rates.

You've seen that UPRO can lose a ton of money and take 17 years to recover. And you expect the long-term treasury returns to make up for that. But I doubt they will make up for any UPRO losses as STRONGLY starting from low interest rates like today.

I mean, your strategy may work, but it is very unlikely to mimic the past 30 years. Your backtest was dependent on conditions that do not exist today.

Again, today's conditions aren't bad. The strategy could work. But it will NOT work in the same way. Your back-test has very little predictive power about what kind of returns you will get going forward.

You are missing my point completely. I make no prediction about whether interest rates will go up or down in the long term, nor how much they will go up or down. Anyone who does should be ignored as a market timer, in the same way as those who predict future stock returns.

Let me state it one more time, as clearly as I can: my strategy rests on the daily correlation between stocks and long term bonds, NOT their medium-to-long term returns. As I am typing this right now, UPRO is down -0.56%, and TMF is up +1.31%. This is the behavior my strategy rests upon, and the behavior I do expect to continue, for risk and market structure reasons. It has nothing to do with long term interest rates or long term stock returns.

Have you thought about including gold in the portfolio? Or some other asset with low correlation to both stocks and bonds? The achilles heel of a stock/long duration bond portfolio is unexpected inflation, because it causes both stocks and bonds to go down in value.

Gold is a good hedge for inflation shocks, not so much for grinding, slowly increasing, "normal" inflation.

Since 1982 inflation has averaged 2.7% / year. I am among those who believe that Paul Volker basically "solved" the problem of inflation 40 years ago. There is a reason the 3-fund portfolio does not include gold.

https://www.economist.com/finance-and-e ... -inflation

[EfficientInvestor]

Agreed. This seems to be the consensus. We need more data then a time period where we already know LT Treasury and Sp500 did very well. Sounds like a broken record by replicating a similar strategy starting after WWII and slowly up to 1970 or so would be good proxy (if any exists) of what is likely to happen next 30 years. A falling interest rate environment is not the same as a slowly rising one in its effect on price of a long term bond.

Yes, totally agreed. I provided daily data series to EfficientInvestor (by PM), starting in Jan-73 for both the S&P 500 and Barclays LT Treasuries indices. I would much prefer to start in the early 60s, but this would require to use simulated data sets (e.g. the bond fund simulator we use for Simba) which would have at best monthly granularity.

One step at a time, let's explore the 1973+ data set first (I ran a coarse simulation myself and although the oil crisis was clearly NOT friendly to such 3x strategy, it doesn't seem to fall apart either -- but I don't master the whole 3x topic enough, so I'd rather wait for EfficientInvestor to run his own numbers). This whole thing IS an intriguing idea, in all fairness, worth exploring in more depth.

Thanks for sending the data! I'll dig into it tonight and over the weekend and hopefully post results at some point tomorrow. I think the 1973-1984 period will be very interesting to look at. I personally prefer the use of gold as an inflation hedge (just in case) and would like to show comparisons of a 40/60 portfolio vs maybe a 35/55/10 portfolio. However, I only have daily gold data going back through 1985. Do you happen to have any data or can you point me in the direction of some daily gold data as well going back to 1973 (or beyond)? Thanks again!

[samsdad]

Agreed. This seems to be the consensus. We need more data then a time period where we already know LT Treasury and Sp500 did very well. Sounds like a broken record by replicating a similar strategy starting after WWII and slowly up to 1970 or so would be good proxy (if any exists) of what is likely to happen next 30 years. A falling interest rate environment is not the same as a slowly rising one in its effect on price of a long term bond.

Yes, totally agreed. I provided daily data series to EfficientInvestor (by PM), starting in Jan-73 for both the S&P 500 and Barclays LT Treasuries indices. I would much prefer to start in the early 60s, but this would require to use simulated data sets (e.g. the bond fund simulator we use for Simba) which would have at best monthly granularity.

One step at a time, let's explore the 1973+ data set first (I ran a coarse simulation myself and although the oil crisis was clearly NOT friendly to such 3x strategy, it doesn't seem to fall apart either -- but I don't master the whole 3x topic enough, so I'd rather wait for EfficientInvestor to run his own numbers). This whole thing IS an intriguing idea, in all fairness, worth exploring in more depth.

Thanks for sending the data! I'll dig into it tonight and over the weekend and hopefully post results at some point tomorrow. I think the 1973-1984 period will be very interesting to look at. I personally prefer the use of gold as an inflation hedge (just in case) and would like to show comparisons of a 40/60 portfolio vs maybe a 35/55/10 portfolio. However, I only have daily gold data going back through 1985. Do you happen to have any data or can you point me in the direction of some daily gold data as well going back to 1973 (or beyond)? Thanks again!

EfficientInvestor, I followed your instructions for ^GSPC and got all the columns back to 1950 except I can't, for the life of me, do the weighting thing at the end. For some reason, Libreoffice doesn't like that formula. Do you have the weighted S&P dataset back to 1950 already? You're right, just the ^GSPC data x 3 understates UPRO dramatically when comparing the two since 2009ish.

[HEDGEFUNDIE]

Some thoughts:

1. I adjusted the time period from 1986 to 1987 in the '87-18 portfolio in column 3, because the TMF fund data provided by EfficientInvestor starts at June 1986 according to PV. To equalize everything, I started it in January 1987 here in Simba's spreadsheet. The CAGR of 8.86 here matches with a 40/60 VFINX (500) and VUSTX (LTT) portfolio in PV for that time period. https://www.portfoliovisualizer.com/bac ... tion2_1=60

2. Comparing the CAGRs of the 40/60 portfolio over those three time periods vs. 100% sp500 and 100% LTT:

Code: Select all

		'50-70	'60-80	'87-18
40/60		6.11%	5.07%	8.86%
100% S&P 500 	12.76%	7.70%	9.87%	
100% LTT	0.98%	2.69%	7.26%

3. Now the question becomes can we take the '50-70 unleveraged data and '60-80 unleveraged data above, and somehow match it to the leveraged data we have from '87-18 via some sort of ratio, even if somewhat imprecise? For example, we know that an unleveraged 40/60 portfolio from '87-18 returned 8.86% CAGR. Had it been leveraged, it would have returned 23.52%. So, can we say that during this time, the leveraged portfolio returned 2.65 times as much? (23.52/8.86=2.65)

If so, can we then extrapolate (or some other word, it's early in the morning), that 2.65 multiplier for the other time periods? So, for the '50-70 data, the unleveraged CAGR of 6.11% would have been 16.22% had it been leveraged? Or the '60-80 unleveraged CAGR of 5.07%--might it have been 13.46% had it been leveraged? Probably not. But someone should be able to come up with an estimate that takes into account such things as the difference in standard deviation, etc. in the data above.

Note that older data for LTT is given only in annual terms, and PV apparently requires at a minimum, monthly returns to come up with a benchmark it'll recognize.

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

But, if you did want to backtest nonetheless, in the absence of daily data, you would need both annual volatility AND annual returns. For each year you would then look up in this table (from the funds' prospectus) what actual return you would have experienced given that intersection of volatility and return.

[Thesaints]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now,

nor the same way as they will in the future !

Just consider that monetary policies are currently the biggest driver in stock AND bond market performance. Doesn't that suggest you that higher correlation might be on the horizon ?

[GrowthSeeker]

Let me state it one more time, as clearly as I can: my strategy rests on the daily correlation between stocks and long term bonds, NOT their medium-to-long term returns.

FWIW, regarding correlation of UPRO and TMF, one can find tools on the internet for checking this. One is at etfreplay , with different look back periods, the shortest being 10 days. It shows a graph of correlation coefficient vs time, but It doesn't provide an average value unfortunately. My eyeball's analogue computer suggests that the average correlation is between -0.25 and -0.5. The graphs look different for different look back periods.

Of course, PV shows this indirectly by showing UPRO's correlation with the stock market to be 1 and TMF's correlation with the stock market to be -0.5. I'm not sure PV is doing this on daily data but that's what I suspect.

But if you have the daily data already in a spreadsheet, you can calculated it for the daily data, I think CORREL() is the Excel function. Maybe that has already been stated here, but I did not see it.

[rmelvey]

Gold is a good hedge for inflation shocks, not so much for grinding, slowly increasing, "normal" inflation.

Since 1982 inflation has averaged 2.7% / year. I am among those who believe that Paul Volker basically "solved" the problem of inflation 40 years ago. There is a reason the 3-fund portfolio does not include gold.

https://www.economist.com/finance-and-e ... -inflation

Yeah I meant inflation shocks (supply curve shifting left specifically). The three fund portfolio bonds have much shorter duration, so I don't think the omission of hard assets is as relevant as it is in your proposed portfolio. If you chose to ignore that risk then i guess that's fine? But it's definitely the achilles heel of your portfolio.

[staythecourse]

Agreed. This seems to be the consensus. We need more data then a time period where we already know LT Treasury and Sp500 did very well. Sounds like a broken record by replicating a similar strategy starting after WWII and slowly up to 1970 or so would be good proxy (if any exists) of what is likely to happen next 30 years. A falling interest rate environment is not the same as a slowly rising one in its effect on price of a long term bond.

Yes, totally agreed. I provided daily data series to EfficientInvestor (by PM), starting in Jan-73 for both the S&P 500 and Barclays LT Treasuries indices. I would much prefer to start in the early 60s, but this would require to use simulated data sets (e.g. the bond fund simulator we use for Simba) which would have at best monthly granularity.

One step at a time, let's explore the 1973+ data set first (I ran a coarse simulation myself and although the oil crisis was clearly NOT friendly to such 3x strategy, it doesn't seem to fall apart either -- but I don't master the whole 3x topic enough, so I'd rather wait for EfficientInvestor to run his own numbers). This whole thing IS an intriguing idea, in all fairness, worth exploring in more depth.

Thanks for sending the data! I'll dig into it tonight and over the weekend and hopefully post results at some point tomorrow. I think the 1973-1984 period will be very interesting to look at. I personally prefer the use of gold as an inflation hedge (just in case) and would like to show comparisons of a 40/60 portfolio vs maybe a 35/55/10 portfolio. However, I only have daily gold data going back through 1985. Do you happen to have any data or can you point me in the direction of some daily gold data as well going back to 1973 (or beyond)? Thanks again!

Much appreciated. At this point I think we all agree (collective) the idea presented and the ability to execute (available etf, trading platform, the right "envelope", i.e. roth ira) is sound. The real question is finding out in less ideal situation what would have happened. The 1970's is one with it ending in super high inflation and another is starting in late 90's at this same high levels of valuation followed by a flat 2000's AND incurring a 2008 debacle in it is another one.

Again when we start talking gold the PP comes into play. How about running the numbers using a 3xleveraged PP as well. As far as I know there is a 3x fund for sp500+ LTT+ gold. Seems Mr. Browne may have been onto "risk parity" before there was a term for it? Maybe he is the father of risk parity?


Good luck.

[unclescrooge]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

But, if you did want to backtest nonetheless, in the absence of daily data, you would need both annual volatility AND annual returns. For each year you would then look up in this table (from the funds' prospectus) what actual return you would have experienced given that intersection of volatility and return.

So if the index rises 10% and volatility rises 100%, the fund's return is -93.4%? Am I interpreting this correctly?

[HEDGEFUNDIE]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

But, if you did want to backtest nonetheless, in the absence of daily data, you would need both annual volatility AND annual returns. For each year you would then look up in this table (from the funds' prospectus) what actual return you would have experienced given that intersection of volatility and return.

So if the index rises 10% and volatility rises 100%, the fund's return is -93.4%? Am I interpreting this correctly?

Yes, but:

1. Volatility "is" 100% for that year not "rises" 100%.
2. Pretty sure annual volatility of the S&P 500 has never come close to 100%, or even 50%. Its average annual volatility since 1976 has been 15%. Between Jan 2007 to Dec 2009 (pretty much the height of volatility), the S&P 500 annual volatility was 20%.

For long Treasuries average annual volatility has been 10% since 1986.

[unclescrooge]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

But, if you did want to backtest nonetheless, in the absence of daily data, you would need both annual volatility AND annual returns. For each year you would then look up in this table (from the funds' prospectus) what actual return you would have experienced given that intersection of volatility and return.

So if the index rises 10% and volatility rises 100%, the fund's return is -93.4%? Am I interpreting this correctly?

Yes, but:

1. Volatility "is" 100% for that year not "rises" 100%.
2. Pretty sure annual volatility of the S&P 500 has never come close to 100%, or even 50%. Its average annual volatility since 1976 has been 15%. Between Jan 2007 to Dec 2009 (pretty much the height of volatility), the S&P 500 annual volatility was 20%.

For long Treasuries average annual volatility has been 10% since 1986.

Ah, that makes sense.

[aj76er]

This is a bit off topic, but would it be easy to rollover a portion of one's Roth into M1? Perhaps something like:

1. Withdraw $5k from current Roth
2. Open Roth at M1 and deposit $5k (within 60 days)

I assume no penalty or taxes in this situation, correct?

[siamond]

Love the idea! I had thought seriously about doing this but vanguard decided to change my mind as of a few weeks ago...

Exactly, which is why my tracker screenshot is of M1 Finance, not Vanguard. Easiest account migration ever (especially when compared with my experience rolling over past accounts to Vanguard).

Could you guys elaborate on what exactly was the issue that lead you to move from Vanguard to M1 Finance? Tx.

[Time2Quit]

Could you guys elaborate on what exactly was the issue that lead you to move from Vanguard to M1 Finance? Tx.

Vanguard won’t allow you to purchase leveraged ETFs

[HEDGEFUNDIE]

This is a bit off topic, but would it be easy to rollover a portion of one's Roth into M1? Perhaps something like:

1. Withdraw $5k from current Roth
2. Open Roth at M1 and deposit $5k (within 60 days)

I assume no penalty or taxes in this situation, correct?

Yes this should be fine. But I would have them handle the transfer for you. Email them and let them know your plan.

All I had to do was email M1 my Vanguard account statement, one week later the money was transferred and ready to go.

[siamond]

The UPRO data is based on daily returns of VFINX and includes a 1% per year expense ratio (I subtract .01/250 for each day). I initially used ^GSPC as the data source, but forgot that I had switched to using VFINX because it accounts for the dividend. If you only use data from ^GSPC, your proxy will underperform UPRO. For each day, I did a weighted average of the close value (20%) and adjusted close value (80%). I determined the weighting using an iterative process until I found something that closely matched the actual UPRO performance.

Are you sure UPRO uses the S&P 500 total return index and not the price index? I could not find a clear answer on ProShares site. It sort of sounds like you are torturing the data until it confesses.

The best I can tell is that it’s somewhere in the middle. UPRO obtains a broad exposure to the index by putting a large portion of funds into individual stocks or unleveraged ETFs. Then, with a small portion of the funds, they use futures and swap contracts to hit the 3x daily goal. By holding the individual stocks and ETFs, they generate a dividend which is currently around 0.5% per Yahoo. Due to this dividend, UPRO is achieving better performance than just 3x the daily ^GSPC (or daily close of VFINX) performance. However, since they can’t apply leverage to that dividend, you can’t assume they will meet the 3x daily performance of the adjusted close of VFINX. Therefore, the actual performance of UPRO is somewhere in the middle. Enter the use of the weighted average. The 20/80 blend that I am using matches actual performance of UPRO quite well over the last 9+ years, so I feel comfortable using that blend going further in the past. If you want to be more conservative, feel free to just use the ^GSPC data without adjusting for dividend.

Although I can appreciate the explanation, it does seem like force fitting to me. There might be all sorts of other frictions happening which makes the UPRO returns significantly less than the "3x daily" goal (incl. transaction costs). I don't have a better idea though. What do you do with TMF? Same problem, right?

Correct. The best methodology I can think of for TMF is to develop a weighted average of the close and adjusted close values of VUSTX. I don't claim it to be perfect. But for me, personally, it is close enough that it would be adequate enough to project what the performance of TMF would have been prior to its inception.

Back to this... Over the 2010-2018 time periods, the VFINX/VTI dividends (~2% annualized) have been WAY higher than the UPRO dividends (~0.2% annualized). 0.5% was UPRO 2018, but this was significantly higher than the previous years (actually, UPRO didn't distribute ANY dividends in 2017, which is weird). This doesn't seem quite consistent with your 20/80 blend formula, does it?

EDIT: located and read your backtest blog entry. Which clarified what you did for TMF. Yeah, this is really curve-fitting on less than a decade of data...

[staythecourse]

This is a bit off topic, but would it be easy to rollover a portion of one's Roth into M1? Perhaps something like:

1. Withdraw $5k from current Roth
2. Open Roth at M1 and deposit $5k (within 60 days)

I assume no penalty or taxes in this situation, correct?

I do believe partial roth rollovers are allowed. I would just call the accepting firm (M1 finance) and complete the paperwork. They will contact Vanguard. On the standard paperwork it will say which portion of your IRA you want transfered, i.e. all or only a part. If it is either you can specify in kind or liquidation as long as the accepting firm accept either way.

Good luck.

[HEDGEFUNDIE]

The entire premise of my strategy rests upon the idea that when one asset collapses, the other asset rises sharply. If you don’t believe this will happen reliably going forward, you should not undertake this strategy. I am comfortable on this point, both because of historical returns and the underlying risk story of why this occurs.

Again, you are only looking at a period of time where long-term bonds were basically in a bull market 90% of the time with falling interest rates.

You've seen that UPRO can lose a ton of money and take 17 years to recover. And you expect the long-term treasury returns to make up for that. But I doubt they will make up for any UPRO losses as STRONGLY starting from low interest rates like today.

I mean, your strategy may work, but it is very unlikely to mimic the past 30 years. Your backtest was dependent on conditions that do not exist today.

Again, today's conditions aren't bad. The strategy could work. But it will NOT work in the same way. Your back-test has very little predictive power about what kind of returns you will get going forward.

Another example for you. August 2000 to April 2010, the “lost decade” for stocks that included two market crashes.

0% total return for the S&P 500.

My strategy would have returned 10.7% CAGR over the same time period.

And before you ask, 30 year Treasuries went from 5.7% to 4.7% during this period, hardly a “bond bull market”.

This is what I mean when I say that this strategy has very little to do with returns, and everything to do with correlations.

[AlohaJoe]

2. Pretty sure annual volatility of the S&P 500 has never come close to 100%, or even 50%. Its average annual volatility since 1976 has been 15%. Between Jan 2007 to Dec 2009 (pretty much the height of volatility), the S&P 500 annual volatility was 20%.

Volatility was 60% during the Great Depression

http://schwert.ssb.rochester.edu/volatility.htm

But it is a bit of an economic mystery why it was so high.

One Federal Reserve paper on the subject wrote

A convincing explanation of why stock volatility
was so high during the Great Depression has eluded scholars

[HEDGEFUNDIE]

2. Pretty sure annual volatility of the S&P 500 has never come close to 100%, or even 50%. Its average annual volatility since 1976 has been 15%. Between Jan 2007 to Dec 2009 (pretty much the height of volatility), the S&P 500 annual volatility was 20%.

Volatility was 60% during the Great Depression

http://schwert.ssb.rochester.edu/volatility.htm

But it is a bit of an economic mystery why it was so high.

One Federal Reserve paper on the subject wrote

A convincing explanation of why stock volatility
was so high during the Great Depression has eluded scholars

Ok, well if we hit the Second Great Depression and this strategy goes to zero, everyone can blame me.

I’m guessing we’ll all have bigger problems to worry about though...

[DiploInvestor]

I can’t give a thorough analysis of the strategy right now, but I can say that this is something I have been doing (in slightly different form) since early 2017. I started with $10k from my Roth IRA using a 60/40 split between the triple leveraged TQQQ (based on Nasdaq 100) and AGG as my safe bond fund at Fidelity. In 2017, my stock fund more than doubled - from $6 to $13k. I rebalance quarterly using a value averaging strategy (see, e.g., Jason Kelly’s 9-sig plan). After 2017, given the confidence I have in the plan and the acceptance of the obvious risks, I placed my entire Roth IRA into this strategy, which means I have about $100k in it right now - about 8% of my total portfolio. It’s already up about 19% for the year. (My and my wife’s TSP accounts and her IRA, and our taxable accounts, follow more sedate strategies: not leveraged, but also mostly value averaged). The 3x leverage in my IRA has been fun to watch, working well, and satisfies a visceral need to be a little more hands on in the market (quarterly rebalancing) without messing with our primary investment vehicles. Using a Nasdaq-based fund hopefully offers some diversity from my other investments that are based on S&P500 and Dow Jones completion index (DWCPF - the TSP “S” fund). Luckily we have the benefit of a healthy TSP and decent pensions to look forward to, so it makes taking risks in my IRA less stressful.

[HEDGEFUNDIE]

I can’t give a thorough analysis of the strategy right now, but I can say that this is something I have been doing (in slightly different form) since early 2017. I started with $10k from my Roth IRA using a 60/40 split between the triple leveraged TQQQ (based on Nasdaq 100) and AGG as my safe bond fund at Fidelity. In 2017, my stock fund more than doubled - from $6 to $13k. I rebalance quarterly using a value averaging strategy (see, e.g., Jason Kelly’s 9-sig plan). After 2017, given the confidence I have in the plan and the acceptance of the obvious risks, I placed my entire Roth IRA into this strategy, which means I have about $100k in it right now - about 8% of my total portfolio. It’s already up about 19% for the year. (My and my wife’s TSP accounts and her IRA, and our taxable accounts, follow more sedate strategies: not leveraged, but also mostly value averaged). The 3x leverage in my IRA has been fun to watch, working well, and satisfies a visceral need to be a little more hands on in the market (quarterly rebalancing) without messing with our primary investment vehicles. Using a Nasdaq-based fund hopefully offers some diversity from my other investments that are based on S&P500 and Dow Jones completion index (DWCPF - the TSP “S” fund). Luckily we have the benefit of a healthy TSP and decent pensions to look forward to, so it makes taking risks in my IRA less stressful.

My concern with using QQQ is that it has much higher volatility than the S&P 500 (24.3% annual volatility vs 14.6% over past 20 years). As the table above from the prospectus shows, the higher the volatility, the higher the chance of getting wiped out. This risk is magnified for you as you’re not holding a leveraged counterweight asset.

As an aside, this is also why I don’t diversify the equity portion of the strategy into small caps, emerging markets, etc. I know 3x funds exist for all those, but they all would add volatility, and would make risk parity harder to achieve.

[fennewaldaj]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

I think it is of interest to see how badly the strategy performs with a higher correlation between stocks and long bonds and a bad decade for both in general. Seems useful to have an idea of how badly this would go if your assumptions break down. I could affect how much one is willing to put in this strategy. It could even tell you to put more in it if turns out to not be too bad.

[FireProof]

Maybe the rebalancing saves things, but I've always seen significant volatility loss over multiple market cycles when I've checked leveraged equity ETFs.

For example, DZK is a 3x leveraged version of EFA (developed markets - a different index of VEA, not particularly volatile).

As of December 31, the unleveraged index has had a respectable 10-year CAGR of 5.84%:
https://www.ishares.com/us/products/239 ... i-eafe-etf

The 3x leveraged index, on the other hand, had a CAGR of only 2.31%
http://www.direxioninvestments.com/prod ... ull-3x-etf

So that's a loss of some 15% CAGR compared to the triple "expected return." Emerging markets and small cap actually had even more loss over the 10 years from 3x leverage (around 21 and 22% CAGR), due to greater volatility, although obviously they have higher volatility than UPRO as well. SPXL, the clone of UPRO, lost around 9.5% CAGR for 10 year trailing returns, despite its very consistent rise.

Daily leveraged bonds perform fine, on the other hand.

And there are increasing options for monthly and quarterly leveraged equity ETNs, which should achieve much closer to the expected return.

[jminv]

Maybe the rebalancing saves things, but I've always seen significant volatility loss over multiple market cycles when I've checked leveraged equity ETFs.

For example, DZK is a 3x leveraged version of EFA (developed markets - a different index of VEA, not particularly volatile).

As of December 31, the unleveraged index has had a respectable 10-year CAGR of 5.84%:
https://www.ishares.com/us/products/239 ... i-eafe-etf

The 3x leveraged index, on the other hand, had a CAGR of only 2.31%
http://www.direxioninvestments.com/prod ... ull-3x-etf

So that's a loss of some 15% CAGR compared to the triple "expected return." Emerging markets and small cap actually had even more loss over the 10 years from 3x leverage (around 21 and 22% CAGR), due to greater volatility, although obviously they have higher volatility than UPRO as well. SPXL, the clone of UPRO, lost around 9.5% CAGR for 10 year trailing returns, despite its very consistent rise.

Daily leveraged bonds perform fine, on the other hand.

And there are increasing options for daily and quarterly leveraged equity ETNs, which should achieve much closer to the expected return.

It's the rebalancing. If you don't rebalance, it wouldn't work nearly as well. You'd end up with around 10% of the final value described in the backtesting. What's being proposed here is basically a risk parity portfolio with leverage then applied. In case you were wondering why he chose quarterly rebalancing, it is because quarterly rebalancing gives the highest return in this data set.

As to the closer daily ETN, using one of these with this strategy could lead to the note losing 100% so it would not be as attractive. Plus the note versus an etf has other undesirable risks.

[Valuethinker]

2. Pretty sure annual volatility of the S&P 500 has never come close to 100%, or even 50%. Its average annual volatility since 1976 has been 15%. Between Jan 2007 to Dec 2009 (pretty much the height of volatility), the S&P 500 annual volatility was 20%.

Volatility was 60% during the Great Depression

http://schwert.ssb.rochester.edu/volatility.htm

But it is a bit of an economic mystery why it was so high.

One Federal Reserve paper on the subject wrote

A convincing explanation of why stock volatility
was so high during the Great Depression has eluded scholars

I cannot check the links on this device. Is the quote from the study?

It seems obvious from an intuitive viewpoint. An enormous shock changed market participant behaviour

Just as our generation has "learned" the "lesson" that you always stay in stocks and buy and hold

So those who experienced 1929 learned that you try to get out when the market is going down because you can lose 80 per cent of your money.

And the macroeconomic volatility of the 1930s was extreme. Just devastating. Not to mention the political volatility.

Socialism was feared in 1930s America by just about everyone with money. FDR was labelled a dictatorial communist.

I would argue that Fascism was a greater risk. Huey Long (And Father Coughlin) in particular. But that too would have brought massive state intervention in the economy.

There were lots of scheme to print money. Serious political schemes. The hard money standard of gold was abolished.

It seems both from a domestic and international viewpoint the 1930s was a time of exceptional instability. In the post-war era the 1970s was perhaps the only real rival.

[vineviz]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

I’m disappointed at how poorly understood this is despite repeated reminders, like yours, that pre-1982 long term Treasury returns are not comparable to post-1982 returns. And the reason has nothing to do with inflation or interest rates.

Not only was the Fed actively suppressing long term rates, but the instruments were callable as well.

[samsdad]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

I’m disappointed at how poorly understood this is despite repeated reminders, like yours, that pre-1982 long term Treasury returns are not comparable to post-1982 returns. And the reason has nothing to do with inflation or interest rates.

Not only was the Fed actively suppressing long term rates, but the instruments were callable as well.

1. I don’t know anything about this area.
2. Things could change in the future.
3. Therefore, I’d like to know how this strategy worked when things were different. See also no. 2, above.
4. I think it’s important to keep in mind that every single investing strategy has its weaknesses. No strategy ever mentioned on this forum is exempt. That doesn’t mean we don’t invest till we find the perfect strategy. If there was one, that’d be the only one before long.

[staythecourse]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

I’m disappointed at how poorly understood this is despite repeated reminders, like yours, that pre-1982 long term Treasury returns are not comparable to post-1982 returns. And the reason has nothing to do with inflation or interest rates.

Not only was the Fed actively suppressing long term rates, but the instruments were callable as well.

Who cares if they were callable. From early 1970's to late 1970's interest rates only increased so it isn't like the treasury was going to call lower coupon bonds for higher coupon bonds. Or am I missing something obvious here?

Even if I am missing something how do you know inflation or interest rates have nothing to do with the strategy if you don't research its effects during times of high interest rates/ inflation? As far as I know we only have one period in modern U.S. history those economics occurred so yes it would be useful. Any person doing research would agree. Now you may be COMPLETELY correct an asterisk needs to be place to suggest it is not the same situation.

No one worth their salt is going to agree on a plan that has one time period analyzed where both the 2 assets did extremely well and just go, "See there you go". No different then every person who goes in the late 1990's, "See just invest in anything that has .com in it and you will be a millionaire".

Mind you this is coming from someone who loves this idea and is eager to delve into it further.

Good luck.

[HEDGEFUNDIE]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

I’m disappointed at how poorly understood this is despite repeated reminders, like yours, that pre-1982 long term Treasury returns are not comparable to post-1982 returns. And the reason has nothing to do with inflation or interest rates.

Not only was the Fed actively suppressing long term rates, but the instruments were callable as well.

1. I don’t know anything about this area.
2. Things could change in the future.
3. Therefore, I’d like to know how this strategy worked when things were different. See also no. 2, above.
4. I think it’s important to keep in mind that every single investing strategy has its weaknesses. No strategy ever mentioned on this forum is exempt. That doesn’t mean we don’t invest till we find the perfect strategy. If there was one, that’d be the only one before long.

If you admit that you don’t know anything about it, shouldn’t your first step be to gain an understanding, instead of wasting time modeling a past that has no predictive power for the future?

As vineviz mentions, consider that Treasuries issued in the 80s and earlier were callable. In other words, if interest rates declined, the Fed could just stop paying interest on the existing bond. So when interest rates did decline, the existing issued Treasuries did not increase in value much, as they would today.

https://www.treasurydirect.gov/indiv/re ... d_call.htm

That is a fundamental difference that would make this strategy behave differently. Interest rates decline during recessions. Stocks drop in these times, and outstanding Treasuries should increase in value. But callable Treasuries would not increase in value, which undercuts the correlations that underpin the strategy.

Let’s flip the situation. If the Fed announced tomorrow that they were bringing back callable bonds, would anyone rely on my 1980s-today backtest? Of course not, and they shouldn’t.

[staythecourse]

I don't necessarily support this effort to push the backtest into the 70s and earlier, as the interest rate regime was very different back then, and long term Treasuries did not necessarily behave with stocks in the same way they do now, given the Fed's change in emphasis after 1982.

I’m disappointed at how poorly understood this is despite repeated reminders, like yours, that pre-1982 long term Treasury returns are not comparable to post-1982 returns. And the reason has nothing to do with inflation or interest rates.

Not only was the Fed actively suppressing long term rates, but the instruments were callable as well.

1. I don’t know anything about this area.
2. Things could change in the future.
3. Therefore, I’d like to know how this strategy worked when things were different. See also no. 2, above.
4. I think it’s important to keep in mind that every single investing strategy has its weaknesses. No strategy ever mentioned on this forum is exempt. That doesn’t mean we don’t invest till we find the perfect strategy. If there was one, that’d be the only one before long.

If you admit that you don’t know anything about it, shouldn’t your first step be to gain an understanding, instead of wasting time modeling a past that has no predictive power for the future?

As vineviz mentions, consider that Treasuries issued in the 80s and earlier were callable. In other words, if interest rates declined, the Fed could just stop paying interest on the existing bond. So when interest rates did decline, the existing issued Treasuries did not increase in value, as they would today. That is a fundamental difference in behavior that would make this strategy behave differently.

https://www.treasurydirect.gov/indiv/re ... d_call.htm

Your link shows the called treasury bonds and the first was in 2000. Is that correct? Which means that there was no called bonds in the 1970's so how would that have affected t bond investors "in the shoes" in 1970's?

Not saying you are not bringing up a good point, BUT if there were non callable at that time it doesn't make a difference to an investor at that time, no? Some clarity on this issue would be great.

I must stay it is starting to sound like you and a few others either already know what the results are during that time period or refuse to open your eyes and review ALL the data. Like I said I love this approach, but like nearly EVERYONE on this site has said it would be nice to see how it would have functioned outside of your one time period. Heck, Fama and French had no problem looking at their Factors across different world markets and time periods to see if the strength of evidence still existed so why would this be an exception?

Good luck.

[HEDGEFUNDIE]

Your link shows the called treasury bonds and the first was in 2000. Is that correct? Which means that there was no called bonds in the 1970's so how would that have affected t bond investors "in the shoes" in 1970's?

Not saying you are not bringing up a good point, BUT if there were non callable at that time it doesn't make a difference to an investor at that time, no? Some clarity on this issue would be great.

I must stay it is starting to sound like you and a few others either already know what the results are during that time period or refuse to open your eyes and review ALL the data. Like I said I love this approach, but like nearly EVERYONE on this site has said it would be nice to see how it would have functioned outside of your one time period. Heck, Fama and French had no problem looking at their Factors across different world markets and time periods to see if the strength of evidence still existed so why would this be an exception?

Good luck.

I have added some detail to my post on callable treasuries. The link shows the last callable treasuries that are still around, not the first ones that were ever issued.

Why do we do backtests? To use the past to guide the future. If the past has nothing to do with the future, why would we do the backtest? People have criticized the 1980s-present backtest by saying it was a huge bond bull market that is unlikely to repeat into the future. And in the same breath they want to backtest into a time when Treasuries were callable and long term yields were actively suppressed? Which is even less likely to recur than another bond bull market! How is this not a contradiction?

[GrowthSeeker]

Gedanken experiment:

Let’s flip the situation. If the Fed announced tomorrow that they were bringing back callable bonds, would anyone rely on my 1980s-today backtest? Of course not, and they shouldn’t.

Theoretically, if the Fed announced tomorrow that they were bringing back callable bonds, how would the markets respond, and more specifically, how do you think the strategy being discussed would respond?
If the answer is that it would hurt the strategy, what are the odds one could get out quickly enough?

[samsdad]

Let’s flip the situation. If the Fed announced tomorrow that they were bringing back callable bonds, would anyone rely on my 1980s-today backtest? Of course not, and they shouldn’t.

That’s the point. Let’s see how this strategy worked in the callable-bond era, just in case our overlords decide to change the rules again.

Listen, I’m thinking about basically doubling down to $10k. If I threw $10k out the window tomorrow would it matter in the long run, all things being equal? Nope. That’s the equivalent of not maxing out your 401K for 6 months. I’ve done that before and am still here. I might have to work another year, etc.

On the other hand, if I wanted to continue to put some into this strategy ever year for awhile, it’d be nice to know how this worked in different periods, with the understanding that nothing’s guaranteed, even based on past similiarities.

It appears that at some point the data becomes more granular and that we don’t have daily data. Well, that’s life. Some people wouldn’t be convinced if we did have daily data. Still, I think data from the “dark ages” would be useful, especially starting in the 1950s.

[HomerJ]

The entire premise of my strategy rests upon the idea that when one asset collapses, the other asset rises sharply. If you don’t believe this will happen reliably going forward, you should not undertake this strategy. I am comfortable on this point, both because of historical returns and the underlying risk story of why this occurs.

Again, you are only looking at a period of time where long-term bonds were basically in a bull market 90% of the time with falling interest rates.

You've seen that UPRO can lose a ton of money and take 17 years to recover. And you expect the long-term treasury returns to make up for that. But I doubt they will make up for any UPRO losses as STRONGLY starting from low interest rates like today.

I mean, your strategy may work, but it is very unlikely to mimic the past 30 years. Your backtest was dependent on conditions that do not exist today.

Again, today's conditions aren't bad. The strategy could work. But it will NOT work in the same way. Your back-test has very little predictive power about what kind of returns you will get going forward.

Another example for you. August 2000 to April 2010, the “lost decade” for stocks that included two market crashes.

0% total return for the S&P 500.

My strategy would have returned 10.7% CAGR over the same time period.

And before you ask, 30 year Treasuries went from 5.7% to 4.7% during this period, hardly a “bond bull market”.

This is what I mean when I say that this strategy has very little to do with returns, and everything to do with correlations.

Umm.. 5.7% dividend and with dropping interest rates, LT bonds returned 7.5% over those 10 years. That is a HUGE part of your 10.7% return over those 10 years.

I'm not saying your strategy won't work... I'm saying the RETURN of 23% a year is unlikely to repeat. You may still beat the market, you may still become quite rich... But I wouldn't count on $100k turning into $10 million.

The conditions aren't exactly the same.

[staythecourse]

Your link shows the called treasury bonds and the first was in 2000. Is that correct? Which means that there was no called bonds in the 1970's so how would that have affected t bond investors "in the shoes" in 1970's?

Not saying you are not bringing up a good point, BUT if there were non callable at that time it doesn't make a difference to an investor at that time, no? Some clarity on this issue would be great.

I must stay it is starting to sound like you and a few others either already know what the results are during that time period or refuse to open your eyes and review ALL the data. Like I said I love this approach, but like nearly EVERYONE on this site has said it would be nice to see how it would have functioned outside of your one time period. Heck, Fama and French had no problem looking at their Factors across different world markets and time periods to see if the strength of evidence still existed so why would this be an exception?

Good luck.

I have added some detail to my post on callable treasuries. The link shows the last callable treasuries that are still around, not the first ones that were ever issued.

Why do we do backtests? To use the past to guide the future. If the past has nothing to do with the future, why would we do the backtest? People have criticized the 1980s-present backtest by saying it was a huge bond bull market that is unlikely to repeat into the future. And in the same breath they want to backtest into a time when Treasuries were callable and long term yields were actively suppressed? How is this not a contradiction?

Do we have a list of when the first treasuries were called? By the link it sounds like it didn't happen before 2000, is that true?

You may be completely correct or not on the other aspects. Like any research the first step is collecting as much data as possible. Then analyzing it. Then figure out if there are measurable reasons why the results are different. Then see if the risk are still outweighed by the pluses. There is no way of knowing what will or won't affect this plan without seeing how it would have performed in different time points, i.e. different economic situations. Are they perfect? Of course not. Just like today's climate does not have to be the same going forward. What prevents the treasury of changing the policy and start to reissue callable bonds?

Again I am really interested in this idea just want to see more info. One as you mentioned is the flat 2000's. Good to know it did well as it started right before with REALLY high valuations just like we have now and then a flat period of returns which includes a complete meltdown. That is very reassuring.

Good luck.

[HomerJ]

Ok, well if we hit the Second Great Depression and this strategy goes to zero, everyone can blame me.

I’m guessing we’ll all have bigger problems to worry about though...

Not me, I'm prepared for a Second Great Depression.

And actually, you could be too, since you're only betting 15% of your money on this strategy.