35/65 portfolio during stock market crashes, 1871-2021

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
namajones
Posts: 778
Joined: Sat Aug 08, 2020 12:41 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by namajones »

Ben Mathew wrote: Tue May 11, 2021 7:31 pm I use Shiller's dataset to see how a 35/65 portfolio weathered stock market crashes. Shiller's data covers 150 years from 1871 to 2021. I converted Shiller's monthly data to annual data before doing the analysis.
This is very interesting, and thank you for doing that.

I wonder whether you would mind substituting the TSP's G Fund for bonds in your analysis. Just curious to see what a bond with no nominal downside would do in such a portfolio over time.

https://www.tsp.gov/funds-individual/g-fund/
namajones
Posts: 778
Joined: Sat Aug 08, 2020 12:41 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by namajones »

SquawkIdent wrote: Wed May 12, 2021 6:13 am Thanks OP for doing all that work.

Enlightening...

Maybe why Wellesley (VWINX and VWIAX) is so popular.
Wellesley appears to be at ~40 percent equities now. Not much difference, I know.

https://investor.vanguard.com/mutual-fu ... olio/vwinx
User avatar
SquawkIdent
Posts: 929
Joined: Tue Dec 23, 2008 6:14 pm
Location: Planet Earth

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by SquawkIdent »

namajones wrote: Sat Jun 05, 2021 11:31 am
SquawkIdent wrote: Wed May 12, 2021 6:13 am Thanks OP for doing all that work.

Enlightening...

Maybe why Wellesley (VWINX and VWIAX) is so popular.
Wellesley appears to be at ~40 percent equities now. Not much difference, I know.

https://investor.vanguard.com/mutual-fu ... olio/vwinx
As long as it’s good with them, it’s good with me. I paying them to keep me out of trouble. :sharebeer
JBTX
Posts: 11227
Joined: Wed Jul 26, 2017 12:46 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by JBTX »

Ben Mathew wrote: Sat Jun 05, 2021 10:41 am
JBTX wrote: Fri Jun 04, 2021 11:24 pm The problem with those who may assert stocks always best bonds over 30 years so one should be 100% stocks is after a certain point the investment time frame becomes less than 30 years.
The investor who believes that stocks always beat bonds over 30 years should, at the 25 year mark, feel that if stocks have underperformed over the last 25 years, then it should do really well over the next 5 years (to catch up and beat bonds). Conversely, if stocks have done well over the last 25 years, the should feel that bonds won't do so well over the next five years. So it implies market timing.

The idea that stocks can't be beaten over 30 years when it can be beaten over 5 years requires significant mean reversion in stocks. Studies have found mean reversion in stock returns, though I don't think the results are robust enough to hang one's hat on. I haven't calculated the probabilities, but my guess is that small or moderate mean reversion would still not drive that the probability of stocks underperforming bonds over 30 years to negligibly low levels. You would likely need high levels of mean reversion.

"No mean reversion" is the more conservative assumption when it comes to financial planning--you may overestimate future uncertainty, but that's better than underestimating it. It also leads to simpler strategies that don't require market timing. So that's the assumption I favor in my financial planning--it's simple and conservative.
I think we agree here. Mean reversion is likely, but relying on it is active investing, and it's timing is by no means reliable. One only has to look at Japan to see that.

But it goes further than that, if you have less than five years left before you need the money, in no rational risk aversion model does it make sense to be 100% stocks. At that point the prior 25 years are irrelevant.

This is why typical retirement glidpaths gradually slope downwards, and I bristle when I see proposed upward sloping retirement glide paths based upon historical US data that shows a certain high equity allocation ends up working out (with the exception that an early retirement pre social security retirement may merit upward sloping prior to SS).
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

namajones wrote: Sat Jun 05, 2021 11:22 am
Ben Mathew wrote: Tue May 11, 2021 7:31 pm I use Shiller's dataset to see how a 35/65 portfolio weathered stock market crashes. Shiller's data covers 150 years from 1871 to 2021. I converted Shiller's monthly data to annual data before doing the analysis.
This is very interesting, and thank you for doing that.

I wonder whether you would mind substituting the TSP's G Fund for bonds in your analysis. Just curious to see what a bond with no nominal downside would do in such a portfolio over time.

https://www.tsp.gov/funds-individual/g-fund/
Shiller has data only for 10 year Treasury bonds. Nominal returns were never negative for these bonds, so the "no nominal downside" analysis can still be done with this data. The main difference would be to keep the bond for 10 years so they collect the positive nominal yield till maturity. Then we could just use the interest as the return, and not worry about price changes. You can get a sense of the difference between bonds held till maturity vs. rolled over continuously in the figure in this post. The "coupon only" series shows the return from holding 10 year bonds to maturity.
Total Portfolio Allocation and Withdrawal (TPAW)
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

JBTX wrote: Sat Jun 05, 2021 2:12 pm
Ben Mathew wrote: Sat Jun 05, 2021 10:41 am
JBTX wrote: Fri Jun 04, 2021 11:24 pm The problem with those who may assert stocks always best bonds over 30 years so one should be 100% stocks is after a certain point the investment time frame becomes less than 30 years.
The investor who believes that stocks always beat bonds over 30 years should, at the 25 year mark, feel that if stocks have underperformed over the last 25 years, then it should do really well over the next 5 years (to catch up and beat bonds). Conversely, if stocks have done well over the last 25 years, the should feel that bonds won't do so well over the next five years. So it implies market timing.

The idea that stocks can't be beaten over 30 years when it can be beaten over 5 years requires significant mean reversion in stocks. Studies have found mean reversion in stock returns, though I don't think the results are robust enough to hang one's hat on. I haven't calculated the probabilities, but my guess is that small or moderate mean reversion would still not drive that the probability of stocks underperforming bonds over 30 years to negligibly low levels. You would likely need high levels of mean reversion.

"No mean reversion" is the more conservative assumption when it comes to financial planning--you may overestimate future uncertainty, but that's better than underestimating it. It also leads to simpler strategies that don't require market timing. So that's the assumption I favor in my financial planning--it's simple and conservative.
I think we agree here. Mean reversion is likely, but relying on it is active investing, and it's timing is by no means reliable. One only has to look at Japan to see that.

But it goes further than that, if you have less than five years left before you need the money, in no rational risk aversion model does it make sense to be 100% stocks. At that point the prior 25 years are irrelevant.

This is why typical retirement glidpaths gradually slope downwards, and I bristle when I see proposed upward sloping retirement glide paths based upon historical US data that shows a certain high equity allocation ends up working out (with the exception that an early retirement pre social security retirement may merit upward sloping prior to SS).
Yes, the standard utility maximization model with CRRA utility and independent returns (so no mean reversion) yields a fixed asset allocation as the optimal strategy during retirement if there are no pensions. I'm not sure what's causing some empirical studies to conclude that upward sloping glide paths can help with sequence of return risk. The math in the standard model says that upward sloping glidepaths will make things worse. I'll have to look into the data for this at some point.

Agree that with a pension starting late, you can get upward sloping glide paths before the pension begins as the bonds funding the gap get consumed.
Total Portfolio Allocation and Withdrawal (TPAW)
JBTX
Posts: 11227
Joined: Wed Jul 26, 2017 12:46 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by JBTX »

Ben Mathew wrote: Sat Jun 05, 2021 2:39 pm
JBTX wrote: Sat Jun 05, 2021 2:12 pm
Ben Mathew wrote: Sat Jun 05, 2021 10:41 am
JBTX wrote: Fri Jun 04, 2021 11:24 pm The problem with those who may assert stocks always best bonds over 30 years so one should be 100% stocks is after a certain point the investment time frame becomes less than 30 years.
The investor who believes that stocks always beat bonds over 30 years should, at the 25 year mark, feel that if stocks have underperformed over the last 25 years, then it should do really well over the next 5 years (to catch up and beat bonds). Conversely, if stocks have done well over the last 25 years, the should feel that bonds won't do so well over the next five years. So it implies market timing.

The idea that stocks can't be beaten over 30 years when it can be beaten over 5 years requires significant mean reversion in stocks. Studies have found mean reversion in stock returns, though I don't think the results are robust enough to hang one's hat on. I haven't calculated the probabilities, but my guess is that small or moderate mean reversion would still not drive that the probability of stocks underperforming bonds over 30 years to negligibly low levels. You would likely need high levels of mean reversion.

"No mean reversion" is the more conservative assumption when it comes to financial planning--you may overestimate future uncertainty, but that's better than underestimating it. It also leads to simpler strategies that don't require market timing. So that's the assumption I favor in my financial planning--it's simple and conservative.
I think we agree here. Mean reversion is likely, but relying on it is active investing, and it's timing is by no means reliable. One only has to look at Japan to see that.

But it goes further than that, if you have less than five years left before you need the money, in no rational risk aversion model does it make sense to be 100% stocks. At that point the prior 25 years are irrelevant.

This is why typical retirement glidpaths gradually slope downwards, and I bristle when I see proposed upward sloping retirement glide paths based upon historical US data that shows a certain high equity allocation ends up working out (with the exception that an early retirement pre social security retirement may merit upward sloping prior to SS).
Yes, the standard utility maximization model with CRRA utility and independent returns (so no mean reversion) yields a fixed asset allocation as the optimal strategy during retirement if there are no pensions. I'm not sure what's causing some empirical studies to conclude that upward sloping glide paths can help with sequence of return risk. The math in the standard model says that upward sloping glidepaths will make things worse. I'll have to look into the data for this at some point.

Agree that with a pension starting late, you can get upward sloping glide paths before the pension begins as the bonds funding the gap gets consumed.
Prefacing I'm not an expert on swr studies, it seems to me they are making the same fallacy as 30 years 100% stocks always wins, so you should be 100% stocks for 30 years. They run all these historical 30 year swr scenarios, quantify the 30 year probalities, and assert that is the 30 year strategy. The reality is you would need to keep rerunning scenarios, 29 years, then 28 years, eventually approaching 0. At five years out your optimal probability will not be 100% stocks.
namajones
Posts: 778
Joined: Sat Aug 08, 2020 12:41 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by namajones »

SquawkIdent wrote: Sat Jun 05, 2021 1:57 pm

As long as it’s good with them, it’s good with me. I paying them to keep me out of trouble. :sharebeer
Isn't Wellesley actively managed? That always makes me nervous. What if the manager changes and the fund goes downhill? Why not just LifeStrategy Conservative Growth Fund? Same mix of stocks and bonds.
seajay
Posts: 1656
Joined: Sat May 01, 2021 3:26 pm
Contact:

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by seajay »

Ben Mathew wrote: Wed May 12, 2021 11:28 am
Blue456 wrote: Wed May 12, 2021 6:28 am Ben, would it be possible to add 30% or 40% ex-US?
Unfortunately, Shiller does not have ex-US data. I don't think there is any ex-US data that goes that far back. If there is a reliable source, I would be interested as well.
UK somewhat fits, a historic global accounting/law/financials hub for the 53 countries 2.3 billion population Commonwealth. The FT All Share index has the top 100 separated out as the FT100 index, the next 250 (350 in combined total) midcap is more reflective of actual UK as the top 100 (FT100) are typically multinational/global outfits just domiciled in the UK for law/accounting/financials. Subjectively ... for instance since Brexit started in 2016 there's been greater drift (lag). VEU (all world ex US) vs EWU (UK)

Barclays Equity Gilt Studies includes data back to 1896 IIRC, so a possible proxy for World ex US.
seajay
Posts: 1656
Joined: Sat May 01, 2021 3:26 pm
Contact:

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by seajay »

TheoLeo wrote: Tue May 11, 2021 7:54 pm I wonder if a bit of gold would have helped much in these post war episodes where bonds didn't help.
These are the differences figures that I'm seeing ...
Cumulative loss of a 35/65 portfolio during these crashes:

World War I (1916-'20): -38% -> -24.8%
Great Depression (1929-'31): -4% -> -3.9%
1937 Recession (1937-'41): -12% -> -32.9%
Post World War II (1946-'47): -25% -> -13%
Oil Crisis (1973-'74): -25% -> +0.6%
Tech Crash (2000-'02): 3% -> -23.7%
Subprime Crisis (2007-'08): 1% -> -18.6%
That's based on a assumption of straight 35/65 for years prior to 1933 when money and gold where the same, convertible at a fixed rate so might as well have money deposited and earning interest. When not pegged it assumes 67/33 stock/precious-metal (PM) i.e. a 50/50 barbell of stock and PM is two polar opposites that combine in a similar manner to a barbell of short and long dated treasury to a central 'bullet'. So 67/33 stock/gold might be considered as a more volatile form of 33/67 type stock/'bond' holding. PV example

It also assumes that silver (poor mans gold) was held instead of gold between 1933 and 1975, gold only from 1976, due to US illegality of holding/trading investment gold during those years.

It also assumes that 67/33 stock/silver was the preferred choice during 1915 to 1918 inclusive World War 1 years (more relevant if a UK investor).

Relevant data ...

Code: Select all

	Real	
1915	-7.3%	Silver
1916	8.8%	Silver
1917	10.1%	Silver
1918	-0.1%	Silver
		
1933	34.3%	Silver
1934	27.6%	Silver
1935	35.4%	Silver
1936	-30.7%	Silver
1937	0.0%	Silver
1938	-2.3%	Silver
1939	-11.3%	Silver
1940	-10.3%	Silver
1941	-3.9%	Silver
1942	-0.5%	Silver
1943	15.4%	Silver
1944	-7.6%	Silver
1945	9.4%	Silver
1946	46.2%	Silver
1947	-20.2%	Silver
1948	-2.4%	Silver
1949	-2.7%	Silver
1950	0.3%	Silver
1951	12.0%	Silver
1952	-4.4%	Silver
1953	-3.0%	Silver
1954	-1.5%	Silver
1955	8.7%	Silver
1956	-2.9%	Silver
1957	-2.8%	Silver
1958	-2.1%	Silver
1959	-0.5%	Silver
1960	-5.1%	Silver
1961	2.1%	Silver
1962	17.2%	Silver
1963	17.5%	Silver
1964	0.7%	Silver
1965	-4.2%	Silver
1966	-3.1%	Silver
1967	20.2%	Silver
1968	31.8%	Silver
1969	-21.7%	Silver
1970	-7.3%	Silver
1971	-15.0%	Silver
1972	3.4%	Silver
1973	43.9%	Silver
1974	65.4%	Silver
1975	-13.2%	Silver
1976	-8.5%	Gold
1977	14.9%	Gold
1978	25.7%	Gold
1979	100.0%	Gold
1980	2.4%	Gold
1981	-38.1%	Gold
1982	10.7%	Gold
1983	-19.4%	Gold
1984	-22.4%	Gold
1985	2.1%	Gold
1986	17.7%	Gold
1987	19.2%	Gold
1988	-18.8%	Gold
1989	-7.2%	Gold
1990	-8.7%	Gold
1991	-11.3%	Gold
1992	-8.4%	Gold
1993	14.5%	Gold
1994	-4.7%	Gold
1995	-1.5%	Gold
1996	-7.7%	Gold
1997	-22.7%	Gold
1998	-2.4%	Gold
1999	-1.8%	Gold
2000	-8.5%	Gold
2001	-0.8%	Gold
2002	22.7%	Gold
2003	17.7%	Gold
2004	1.3%	Gold
2005	13.9%	Gold
2006	19.5%	Gold
2007	25.3%	Gold
2008	4.8%	Gold
2009	20.7%	Gold
2010	27.4%	Gold
2011	6.4%	Gold
2012	4.8%	Gold
2013	-29.4%	Gold
2014	-2.9%	Gold
2015	-11.3%	Gold
2016	5.8%	Gold
2017	10.5%	Gold
2018	-3.8%	Gold
2019	15.2%	Gold
2020	23.1%	Gold
Thanks Ben (and to Shiller also), a great thread/data-analysis. Interesting to me as I like a split-the-difference. UK based holding 50% weighting to UK FT All Share/10 year Gilts (Treasury) 50/50 that to me is a form of global ex US and domestic bonds; Along with a 50% weighting into 50/50 US stock/gold (so 25% each of UK large cap stock (as a form of 'international ex US), US stock, gold, bonds). For UK all 50 year periods (calendar year granularity) since 1896 with caveats similar to as outlined earlier (except that it was lawful to hold investment gold since 1933 in the UK) I see a 3% PWR (perpetual withdrawal rate where you ended with at least as much inflation adjusted portfolio value as at the start) and where on average (median) a additional 3% real occurred on top. Similar depending upon which side of the pond you are

Without SWR the average was 5%. The reason why 3% SWR + 3% additional real average is greater than the 5% with no SWR average is simply down to 3% SWR more often tends to become a smaller percentage of the ongoing portfolio value over time as the remainder portfolio value pulls ahead of inflation.

Since 1986 a relatively stable upward sloping (log scaled) real gain progression line with a exponential R-squared of 0.988. Near straight lined excepting for around a dip during WW1 years.

A noteworthy point is that even though that was a PWR at times drawdowns were quite substantial, for instance the 50 years started 1906 at the end of 1920 was down to 33% of the inflation adjusted start date value (-66% down in real terms), but that did subsequently continue on to 'recover'. Also the PWR was improved by averaging in, half at the start of year, other half at the end of year, that helped avoid lumping-all-in/starting at the worst possible time.

EDIT/PS ... I should have mentioned that for UK bond data I used 10 year ladder data i.e. not marked to market just the ten year rolling average of ten year treasury yields.
seajay
Posts: 1656
Joined: Sat May 01, 2021 3:26 pm
Contact:

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by seajay »

JBTX wrote: Sat Jun 05, 2021 2:49 pm
JBTX wrote: Fri Jun 04, 2021 11:24 pm The problem with those who may assert stocks always best bonds over 30 years so one should be 100% stocks is after a certain point the investment time frame becomes less than 30 years.
Prefacing I'm not an expert on swr studies, it seems to me they are making the same fallacy as 30 years 100% stocks always wins, so you should be 100% stocks for 30 years. They run all these historical 30 year swr scenarios, quantify the 30 year probalities, and assert that is the 30 year strategy. The reality is you would need to keep rerunning scenarios, 29 years, then 28 years, eventually approaching 0. At five years out your optimal probability will not be 100% stocks.
Click the inflation adjusted tickbox in this link where a 4% SWR is being applied since 2000 to US stock (red line), and the suggestion is a 11% annualised over the next 9 years just to catch-up to 25% each in US/world-ex-US/gold/bonds (blue line) is required in order that the two might have at least compared over 30 years.

Image

A example of a bad sequence of returns. So if stocks periodically hit a bad SoR situation where bonds hold/gain value then the suggestion is that stock gains after deep dives always recover very strongly within 30 years or less. If 20 years in (2020) all-stock was down -70% whilst another asset such as bonds were breaking even, then with just 10 years of the clock remaining stocks would have to annualize 13% real over the next decade (to 2030). Possible, but a remote possibility IMO.

Yes I know, 25% each US/ex-US/gold/bonds isn't a 'bond', subject to what you might consider to be a bond. Some 'bonds' (junk/corporate/whatever) might more reflect that asset allocation. Some bonds are more stock-like, some stocks are more bond-like.
User avatar
SquawkIdent
Posts: 929
Joined: Tue Dec 23, 2008 6:14 pm
Location: Planet Earth

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by SquawkIdent »

namajones wrote: Sat Jun 05, 2021 4:55 pm
SquawkIdent wrote: Sat Jun 05, 2021 1:57 pm

As long as it’s good with them, it’s good with me. I paying them to keep me out of trouble. :sharebeer
Isn't Wellesley actively managed? That always makes me nervous. What if the manager changes and the fund goes downhill? Why not just LifeStrategy Conservative Growth Fund? Same mix of stocks and bonds.
Yes, Wellesley is actively managed. The fees for each fund are similar.

Head on over to the VG and check out the holdings of each fund. While they are both pretty close to 40S/60B, the holdings they have are different. It all depends on what you want your fund to hold.
seajay
Posts: 1656
Joined: Sat May 01, 2021 3:26 pm
Contact:

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by seajay »

seajay wrote: Sun Jun 06, 2021 7:54 am
TheoLeo wrote: Tue May 11, 2021 7:54 pm I wonder if a bit of gold would have helped much in these post war episodes where bonds didn't help.
These are the differences figures that I'm seeing ...
Cumulative loss of a 35/65 portfolio during these crashes:

World War I (1916-'20): -38% -> -24.8%
Great Depression (1929-'31): -4% -> -3.9%
1937 Recession (1937-'41): -12% -> -32.9%
Post World War II (1946-'47): -25% -> -13%
Oil Crisis (1973-'74): -25% -> +0.6%
Tech Crash (2000-'02): 3% -> -23.7%
Subprime Crisis (2007-'08): 1% -> -18.6%
That's based on a assumption of straight 35/65 for years prior to 1933 when money and gold where the same, convertible at a fixed rate so might as well have money deposited and earning interest. When not pegged it assumes 67/33 stock/precious-metal (PM) i.e. a 50/50 barbell of stock and PM is two polar opposites that combine in a similar manner to a barbell of short and long dated treasury to a central 'bullet'. So 67/33 stock/gold might be considered as a more volatile form of 33/67 type stock/'bond' holding. PV example

It also assumes that silver (poor mans gold) was held instead of gold between 1933 and 1975, gold only from 1976, due to US illegality of holding/trading investment gold during those years.

It also assumes that 67/33 stock/silver was the preferred choice during 1915 to 1918 inclusive World War 1 years (more relevant if a UK investor).

Relevant data ...

Code: Select all

	Real	
1915	-7.3%	Silver
1916	8.8%	Silver
1917	10.1%	Silver
1918	-0.1%	Silver
		
1933	34.3%	Silver
1934	27.6%	Silver
1935	35.4%	Silver
1936	-30.7%	Silver
1937	0.0%	Silver
1938	-2.3%	Silver
1939	-11.3%	Silver
1940	-10.3%	Silver
1941	-3.9%	Silver
1942	-0.5%	Silver
1943	15.4%	Silver
1944	-7.6%	Silver
1945	9.4%	Silver
1946	46.2%	Silver
1947	-20.2%	Silver
1948	-2.4%	Silver
1949	-2.7%	Silver
1950	0.3%	Silver
1951	12.0%	Silver
1952	-4.4%	Silver
1953	-3.0%	Silver
1954	-1.5%	Silver
1955	8.7%	Silver
1956	-2.9%	Silver
1957	-2.8%	Silver
1958	-2.1%	Silver
1959	-0.5%	Silver
1960	-5.1%	Silver
1961	2.1%	Silver
1962	17.2%	Silver
1963	17.5%	Silver
1964	0.7%	Silver
1965	-4.2%	Silver
1966	-3.1%	Silver
1967	20.2%	Silver
1968	31.8%	Silver
1969	-21.7%	Silver
1970	-7.3%	Silver
1971	-15.0%	Silver
1972	3.4%	Silver
1973	43.9%	Silver
1974	65.4%	Silver
1975	-13.2%	Silver
1976	-8.5%	Gold
1977	14.9%	Gold
1978	25.7%	Gold
1979	100.0%	Gold
1980	2.4%	Gold
1981	-38.1%	Gold
1982	10.7%	Gold
1983	-19.4%	Gold
1984	-22.4%	Gold
1985	2.1%	Gold
1986	17.7%	Gold
1987	19.2%	Gold
1988	-18.8%	Gold
1989	-7.2%	Gold
1990	-8.7%	Gold
1991	-11.3%	Gold
1992	-8.4%	Gold
1993	14.5%	Gold
1994	-4.7%	Gold
1995	-1.5%	Gold
1996	-7.7%	Gold
1997	-22.7%	Gold
1998	-2.4%	Gold
1999	-1.8%	Gold
2000	-8.5%	Gold
2001	-0.8%	Gold
2002	22.7%	Gold
2003	17.7%	Gold
2004	1.3%	Gold
2005	13.9%	Gold
2006	19.5%	Gold
2007	25.3%	Gold
2008	4.8%	Gold
2009	20.7%	Gold
2010	27.4%	Gold
2011	6.4%	Gold
2012	4.8%	Gold
2013	-29.4%	Gold
2014	-2.9%	Gold
2015	-11.3%	Gold
2016	5.8%	Gold
2017	10.5%	Gold
2018	-3.8%	Gold
2019	15.2%	Gold
2020	23.1%	Gold
Thanks Ben (and to Shiller also), a great thread/data-analysis. Interesting to me as I like a split-the-difference. UK based holding 50% weighting to UK FT All Share/10 year Gilts (Treasury) 50/50 that to me is a form of global ex US and domestic bonds; Along with a 50% weighting into 50/50 US stock/gold (so 25% each of UK large cap stock (as a form of 'international ex US), US stock, gold, bonds). For UK all 50 year periods (calendar year granularity) since 1896 with caveats similar to as outlined earlier (except that it was lawful to hold investment gold since 1933 in the UK) I see a 3% PWR (perpetual withdrawal rate where you ended with at least as much inflation adjusted portfolio value as at the start) and where on average (median) a additional 3% real occurred on top. Similar depending upon which side of the pond you are

Without SWR the average was 5%. The reason why 3% SWR + 3% additional real average is greater than the 5% with no SWR average is simply down to 3% SWR more often tends to become a smaller percentage of the ongoing portfolio value over time as the remainder portfolio value pulls ahead of inflation.

Since 1986 a relatively stable upward sloping (log scaled) real gain progression line with a exponential R-squared of 0.988. Near straight lined excepting for around a dip during WW1 years.

A noteworthy point is that even though that was a PWR at times drawdowns were quite substantial, for instance the 50 years started 1906 at the end of 1920 was down to 33% of the inflation adjusted start date value (-66% down in real terms), but that did subsequently continue on to 'recover'. Also the PWR was improved by averaging in, half at the start of year, other half at the end of year, that helped avoid lumping-all-in/starting at the worst possible time.

EDIT/PS ... I should have mentioned that for UK bond data I used 10 year ladder data i.e. not marked to market just the ten year rolling average of ten year treasury yields.
Digging deeper and setting to use 50/50 stock/bond or 75/25 stock/PM (and 50/50 stock/bond in earlier years ...etc. caveats), and then comparing ....

Image

and the arithmetic/geometric/median/stdev ...etc. are all over the place. So I opted to simply use the compounded bad-times outcomes and they all came out at pretty much the same.

Looks to me broadly that it didn't matter if you opted for 50/50 stock/bonds, or 75/25 stock/gold as a form of 50/50 stock/bond, at least across multiple bad-times. So just pick whichever is the more cost/tax efficient for you in that sense. However overall, across all 150 years, the 75/25 stock/gold choice annualized 1.24% more.

To check whether that was a localized outcome (choice of start/end dates) I measured all 10 year periods and compared those ...

Code: Select all

 Avg 	5.1%	6.4%
Median	5.2%	6.5%
Stdev	3.7%	3.3%
   50/50 S/B    75/25 S/G
and it looks like 50% stock, 50% stock/gold 50/50 as a bond barbell type holding endured comparable bad-times overall outcome to that of 50/50 stock/bond, but provided higher overall rewards (better risk-adjusted reward).
Blue456
Posts: 2152
Joined: Tue Jun 04, 2019 5:46 am

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Blue456 »

qwertyjazz wrote: Sat Jun 05, 2021 11:17 am
Ben Mathew wrote: Fri May 14, 2021 8:55 am
Volando wrote: Thu May 13, 2021 9:09 pm
Ben Mathew wrote: Thu May 13, 2021 7:51 pm But the problem is that this is not a lot of data points. 150 years is still only 5 separate 30 year periods. The profusion of dots you see are coming off the same underlying annual data. Each period is just slightly different from the period before and after it. So maybe bonds will actually outperform once in a while. It simply may not have shown up in our data.
I can see how 150/30 = 5. But in reality wouldn't you have a much larger number of 30 year periods if you delved a little deeper? I believe the chart you shared has 121, 30 year periods. Further is this data only limited to annual data or can we get down to months? If you delved even deeper by going into specific months, say starting from May 1871- 1901 compared to June 1871 - 1901 you'd have even more 30 year periods to evaluate. From my basic understanding of all of this I can imagine that starting at different points in time would yield different results.
Even though there are 121 30 year periods, they are not truly distinct data points. The 1871-1901 period is almost the same at the 1872-1902 period. If the former was a good draw, the latter will almost certainly be a good as well. So they are not independent draws. They are highly correlated (serial correlation). You would have to go out to 1901-1931 to get a truly new data point. In that sense there are only 5 completely independent data points here:

1871-1901
1901-1931
1931-1961
1961-1991
1991-2021

Imagine that these 5 independent draws happened to be good. Now all the other 116 periods will be forced to be exceptionally good on average. Take for example 1920 -1950. It's about two thirds 1901-1931 and one thirds 1931-1961. So it can't be too bad given that the latter two were good.
Volando wrote: Thu May 13, 2021 9:09 pm What would constitute a large enough set of data points to be sufficiently confident in the assertion that a majority(or entirely) stock portfolio will continue to outperform one with less?
There are ways to calculate standard errors and so on with serial correlated data. With enough data points, those standard errors would become tight. I haven't gone down the route of calculating this. The simulations from annual returns are easier to understand and can tell us what we need to know, which is the probability of stocks doing really badly over 30 years.
Volando wrote: Thu May 13, 2021 9:09 pm
Ben Mathew wrote: Thu May 13, 2021 7:51 pm So then is there any other way of estimating the probability the bonds will outperform stocks over 30 years? Indeed there is. If you are willing to assume that the annual returns are independent, you could simulate thousands of paths using the 150 annual draws that we have. And some of those paths will be very very bad for stocks. If you are unlucky enough, you can get stocks doing as badly as the worst year of the great depression every single year. So averaging across all these simulations, we will get some nonzero probability that stocks will do really badly, and bonds helped tremendously. That gives you the reason to hold bonds over long horizons.
Regarding this point I think I see what you're saying. My stats skills are rusty but here goes: There's a whole range of possible outcomes that could have occurred within this time frame and some could have had stocks doing very poorly, and even consistently doing very poorly. Correct? However, would there not be a range in which the majority of our outcomes would fall which are reasonably represented by what we have seen in the historical data? I'm thinking of a probability distribution where the bulk of the outcomes fall within a certain range, while the extreme outcomes fall on the edges. Although it is possible for there to be repetitive poor outcomes for stocks, isn't it reasonable to assume that the behavior of stocks and bonds will continue to result somewhat similarly to what we have already seen over 150 years? The extremes strike me as so remote and would require such profound shifts in our world that I'm not sure any asset allocation would survive very long, making it more academic than practical.
This comes back to the previous point of how many independent 30-year data points are actually there in 150 years. If we go the route of simulating thousands of 30 year returns from annual returns, then we can directly see if the probability of stock underperformance is so unlikely that bonds will not be useful. My guess is that stocks will have a significant probability of very poor performance, and that the average investor will need to hold bonds over long horizons.

Note also that if stocks will definitely outperform bonds over long horizons like 30 years, there would be an arbitrage opportunity in issuing noncallable debt over 30 years and investing in stocks--i.e. people and institutions should be infinitely leveraged over long horizons. That's not what we see, and so that itself suggests that the risk of stock underperformance is non-negligible.
The first limited liability law was written in New York in 1811. So any concept of investing in modern corporations (stocks or bonds) has to date after that. That limits you to 6-7 independent 30 year time frames.
A rule of thumb in stats is for each ten data points you can look at one variable (of course there are issues with that). So one could argue that you do not have enough data to loo at any full retirement. This of course assumes stability in the process you are modeling. Investing in stocks was very different for a small investor 80 years ago let alone 160.
I am confused why you think there were only 6-7 independent 30 year periods. Each year a person retires can be considered independent (to that person) 30 year period.
mr_brightside
Posts: 897
Joined: Sat Oct 17, 2020 3:23 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by mr_brightside »

that graph is a horrible depiction

bottom axis each year is represented as an equal value -- fair enough -- every year is 365 days (okay except leap years...)

but the vertical axis is way off in scale... the first measure is 10, then 100, then 1000, then 10,000 -- but each 'gradient' is depicted by the same physical measurement even though the values are increasing exponentially.

result : the difference in the results of the portfolios looks WAY tighter than it really is.

over 150 years the difference between 4.5% and 6.5% is MASSIVE but the graph conceals it.

--------------------------------------
okwriter
Posts: 242
Joined: Mon Apr 12, 2021 2:00 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by okwriter »

It's just in log scale, nothing "way off" about it. Personally I prefer log scaling because the slope tells you the rate of growth, which makes it easier to compare the 3 portfolios. For example, in the 90s the slopes are similar, meaning bonds grew nearly as fast as stocks. But in the 50s there was a big difference.
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

mr_brightside wrote: Wed Jun 16, 2021 5:46 pm that graph is a horrible depiction

bottom axis each year is represented as an equal value -- fair enough -- every year is 365 days (okay except leap years...)

but the vertical axis is way off in scale... the first measure is 10, then 100, then 1000, then 10,000 -- but each 'gradient' is depicted by the same physical measurement even though the values are increasing exponentially.

result : the difference in the results of the portfolios looks WAY tighter than it really is.

over 150 years the difference between 4.5% and 6.5% is MASSIVE but the graph conceals it.
The goal of that graph was not to show the power of compounding over 150 years, but to identify stock market crashes over the last 150 years.
A log scale is necessary to show each crash on an equal footing. With a log scale a 50% crash would show the same drop regardless of when it happened. Without a log scale, everything becomes dominated by the last few years. It would actually hide the size of the early crashes precisely because of compounding.
Total Portfolio Allocation and Withdrawal (TPAW)
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

Blue456 wrote: Wed Jun 16, 2021 5:25 pm
qwertyjazz wrote: Sat Jun 05, 2021 11:17 am
The first limited liability law was written in New York in 1811. So any concept of investing in modern corporations (stocks or bonds) has to date after that. That limits you to 6-7 independent 30 year time frames.
A rule of thumb in stats is for each ten data points you can look at one variable (of course there are issues with that). So one could argue that you do not have enough data to loo at any full retirement. This of course assumes stability in the process you are modeling. Investing in stocks was very different for a small investor 80 years ago let alone 160.
I am confused why you think there were only 6-7 independent 30 year periods. Each year a person retires can be considered independent (to that person) 30 year period.
It's independence as defined in probability that matters for drawing statistical conclusions. So the fact that each year is new to some retiree does not help.
Total Portfolio Allocation and Withdrawal (TPAW)
Blue456
Posts: 2152
Joined: Tue Jun 04, 2019 5:46 am

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Blue456 »

Ben Mathew wrote: Wed Jun 16, 2021 6:46 pm
Blue456 wrote: Wed Jun 16, 2021 5:25 pm
qwertyjazz wrote: Sat Jun 05, 2021 11:17 am
The first limited liability law was written in New York in 1811. So any concept of investing in modern corporations (stocks or bonds) has to date after that. That limits you to 6-7 independent 30 year time frames.
A rule of thumb in stats is for each ten data points you can look at one variable (of course there are issues with that). So one could argue that you do not have enough data to loo at any full retirement. This of course assumes stability in the process you are modeling. Investing in stocks was very different for a small investor 80 years ago let alone 160.
I am confused why you think there were only 6-7 independent 30 year periods. Each year a person retires can be considered independent (to that person) 30 year period.
It's independence as defined in probability that matters for drawing statistical conclusions. So the fact that each year is new to some retiree does not help.
But for practical reasons it would make a huge difference for someone retiring in 2000s vs 2010s.
qwertyjazz
Posts: 2000
Joined: Tue Feb 23, 2016 3:24 am

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by qwertyjazz »

Ben Mathew wrote: Wed Jun 16, 2021 6:46 pm
Blue456 wrote: Wed Jun 16, 2021 5:25 pm
qwertyjazz wrote: Sat Jun 05, 2021 11:17 am
The first limited liability law was written in New York in 1811. So any concept of investing in modern corporations (stocks or bonds) has to date after that. That limits you to 6-7 independent 30 year time frames.
A rule of thumb in stats is for each ten data points you can look at one variable (of course there are issues with that). So one could argue that you do not have enough data to loo at any full retirement. This of course assumes stability in the process you are modeling. Investing in stocks was very different for a small investor 80 years ago let alone 160.
I am confused why you think there were only 6-7 independent 30 year periods. Each year a person retires can be considered independent (to that person) 30 year period.
It's independence as defined in probability that matters for drawing statistical conclusions. So the fact that each year is new to some retiree does not help.
It is not just in financial economic history but multiple fields where we on the one hand go there is not enough data to make any predictions but then go this seems to be what works. I am not sure of your initial thesis of 35/65 but I still kinda believe it from your analysis. I think I am as a human being destined to not believe in statistics.
G.E. Box "All models are wrong, but some are useful."
AlohaJoe
Posts: 6609
Joined: Mon Nov 26, 2007 1:00 pm
Location: Saigon, Vietnam

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by AlohaJoe »

Ben Mathew wrote: Wed May 12, 2021 11:28 am
Blue456 wrote: Wed May 12, 2021 6:28 am Ben, would it be possible to add 30% or 40% ex-US?
Unfortunately, Shiller does not have ex-US data. I don't think there is any ex-US data that goes that far back. If there is a reliable source, I would be interested as well.
The Jordà-Schularick-Taylor Macrohistory Database covers 18 economies back to 1870. But this historical data is less useful than you'd think, since you don't know how to weight the returns. There is no public database of historical market capitalisation over time. There is a paper that covers it but I don't think the author ever made his data public. Once upon a time siamond was in contact with the author but I'm not sure anything ever came of it.

https://www.macrohistory.net/database/
AlohaJoe
Posts: 6609
Joined: Mon Nov 26, 2007 1:00 pm
Location: Saigon, Vietnam

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by AlohaJoe »

Ben Mathew wrote: Thu May 13, 2021 7:51 pm Stocks outperformed bonds every single time.

So what can we conclude about the probability (P) of bonds outperforming stocks ? If you conclude that P=0 because it never happened in this data, then it makes sense to conclude that for goals with horizon >= 30 years, there is no reason to hold bonds. Stocks will always outperform.

But the problem is that this is not a lot of data points. 150 years is still only 5 separate 30 year periods. The profusion of dots you see are coming off the same underlying annual data.
In his 2012 paper "Are Stocks Riskier Than Bonds? Not If You Assess Risk Like Warren Buffett" Javier Estrada compares stocks & bonds 19 countries to see how often stocks out performed bonds over 20 & 30 years periods, as well as how often each delivered negative real returns over long periods of time.

Image
User avatar
beyou
Posts: 6915
Joined: Sat Feb 27, 2010 2:57 pm
Location: If you can make it there

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by beyou »

JBTX wrote: Fri Jun 04, 2021 11:24 pm The problem with those who may assert stocks always best bonds over 30 years so one should be 100% stocks is after a certain point the investment time frame becomes less than 30 years.
Unless one is planning to pass a portfolio down to the next generation.
Northern Flicker
Posts: 15363
Joined: Fri Apr 10, 2015 12:29 am

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Northern Flicker »

Ben Mathew wrote: Shiller's data covers 150 years from 1871 to 2021....
Stock is S&P 500 including dividends.
The S&P500 was created in 1957.
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

qwertyjazz wrote: Wed Jun 16, 2021 11:15 pm
Ben Mathew wrote: Wed Jun 16, 2021 6:46 pm
Blue456 wrote: Wed Jun 16, 2021 5:25 pm
qwertyjazz wrote: Sat Jun 05, 2021 11:17 am
The first limited liability law was written in New York in 1811. So any concept of investing in modern corporations (stocks or bonds) has to date after that. That limits you to 6-7 independent 30 year time frames.
A rule of thumb in stats is for each ten data points you can look at one variable (of course there are issues with that). So one could argue that you do not have enough data to loo at any full retirement. This of course assumes stability in the process you are modeling. Investing in stocks was very different for a small investor 80 years ago let alone 160.
I am confused why you think there were only 6-7 independent 30 year periods. Each year a person retires can be considered independent (to that person) 30 year period.
It's independence as defined in probability that matters for drawing statistical conclusions. So the fact that each year is new to some retiree does not help.
It is not just in financial economic history but multiple fields where we on the one hand go there is not enough data to make any predictions but then go this seems to be what works. I am not sure of your initial thesis of 35/65 but I still kinda believe it from your analysis. I think I am as a human being destined to not believe in statistics.
I wouldn't say that there is not enough data to make any predictions. In light of the lack of independence of the data, the predictions would simply be more uncertain, and would allow for events that have not happened in the past. Assuming annual returns are independent for example opens the door to another way of estimating the probability that stocks underperform bonds. This method would lead to a non-zero probability that stocks underperform bonds.
Total Portfolio Allocation and Withdrawal (TPAW)
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

AlohaJoe wrote: Wed Jun 16, 2021 11:59 pm
Ben Mathew wrote: Wed May 12, 2021 11:28 am
Blue456 wrote: Wed May 12, 2021 6:28 am Ben, would it be possible to add 30% or 40% ex-US?
Unfortunately, Shiller does not have ex-US data. I don't think there is any ex-US data that goes that far back. If there is a reliable source, I would be interested as well.
The Jordà-Schularick-Taylor Macrohistory Database covers 18 economies back to 1870. But this historical data is less useful than you'd think, since you don't know how to weight the returns. There is no public database of historical market capitalisation over time. There is a paper that covers it but I don't think the author ever made his data public. Once upon a time siamond was in contact with the author but I'm not sure anything ever came of it.

https://www.macrohistory.net/database/
Thanks for the lead. I will explore this.
Total Portfolio Allocation and Withdrawal (TPAW)
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

AlohaJoe wrote: Thu Jun 17, 2021 12:06 am In his 2012 paper "Are Stocks Riskier Than Bonds? Not If You Assess Risk Like Warren Buffett" Javier Estrada compares stocks & bonds 19 countries to see how often stocks out performed bonds over 20 & 30 years periods, as well as how often each delivered negative real returns over long periods of time.

Image
Thanks. Taking an international perspective is always useful. Seems difficult to believe that stocks can underperform bonds over the long term in other countries, but not in the U.S.

I think the US has simply done better than expected. I've heard that Samuelson's economics textbook used to have a graph that showed when the Soviet Union will overtake the US. That the US did so much better than the rest of the world was not a sure thing always. But it did do better. And now, an expectation that it will continue to do better than the rest of the world has become priced into US stock valuations. This has added to the spectacular US historical returns, but does not bode well for future returns. I don't know if it will do worse than the rest of the world, but it at least argues for a normalization. I would be very surprised if the next 150 years of US stock returns looks like the last 150 years--even if US companies continue to do better than the rest of the world.
Total Portfolio Allocation and Withdrawal (TPAW)
User avatar
Topic Author
Ben Mathew
Posts: 2720
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Ben Mathew »

Northern Flicker wrote: Thu Jun 17, 2021 1:30 pm
Ben Mathew wrote: Shiller's data covers 150 years from 1871 to 2021....
Stock is S&P 500 including dividends.
The S&P500 was created in 1957.
Yes, I didn't go into the details of pre-1957 data. It's available on Shiller's data page, which I linked to in the OP. I believe this is the closest we can get to something like the S&P500--basically market cap weighted returns of the largest US companies.
Total Portfolio Allocation and Withdrawal (TPAW)
AlohaJoe
Posts: 6609
Joined: Mon Nov 26, 2007 1:00 pm
Location: Saigon, Vietnam

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by AlohaJoe »

Ben Mathew wrote: Thu Jun 17, 2021 6:12 pm Thanks. Taking an international perspective is always useful. Seems difficult to believe that stocks can underperform bonds over the long term in other countries, but not in the U.S.
It's always tricky to assess international numbers, especially before the 1950s, because there's a ton of nuance happening behind the scenes that make it hard to know what any of it means.

One factor: most countries simply don't have a diversified set of public companies. Even today outside of the US, China, UK, and Japan how many other truly diversified economies could you add to the list? Maybe France and Germany? Most economies are dominated by just two or three sectors. So when it says "Finnish stocks underperformed" it is probably better to read that as a sector under-performing.

Another factor: the World Wars. (Or in Spain's case, its own civil war.) In some cases that meant government bonds performed very poorly as they either outright defaulted or simply had high inflation. But in other cases it meant stocks performed very poorly as an entire nation's factories were blown up.

Also: Almost nobody, even back in decades ago, would have invested 100% in a single European country like Belgium or the Netherlands. Even before the European Union, it was somewhat easier to invest across Europe.

If you look just at the ex-US countries not directly involved in WW1/2 -- Canada, South Africa, Australia, New Zealand, Sweden, Ireland -- then stocks look to "win" well over 95% of the time. But then we've got another confounding factor: notice that all of them (other than Sweden) are Anglo countries. Is there some Secret Sauce (better institutions, maybe?) there that will repeat going forward? Maybe? Or is it just a coincidence of geography and history?
Northern Flicker
Posts: 15363
Joined: Fri Apr 10, 2015 12:29 am

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by Northern Flicker »

Ben Mathew wrote: Thu Jun 17, 2021 6:16 pm
Northern Flicker wrote: Thu Jun 17, 2021 1:30 pm
Ben Mathew wrote: Shiller's data covers 150 years from 1871 to 2021....
Stock is S&P 500 including dividends.
The S&P500 was created in 1957.
Yes, I didn't go into the details of pre-1957 data. It's available on Shiller's data page, which I linked to in the OP. I believe this is the closest we can get to something like the S&P500--basically market cap weighted returns of the largest US companies.
But a much smaller percentage of US GDP was performed in publicly traded companies in the 19th century. 1871-1900 US large caps were a collection mainly of banks and railroads. Sector risk was a problem even with the broad market. It could not be diversified away in the 19th century so holding stocks was riskier but risk that would be diversifiable and uncompensated today would thus have been compensated then. Stocks as an asset class had a higher expected return in 1871 or 1900 than today, but also were riskier.

Once the data is put into a graph, it becomes easy to assume that stocks in different periods of time behaved the same way, but that is not the case.
namajones
Posts: 778
Joined: Sat Aug 08, 2020 12:41 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by namajones »

Ben Mathew wrote: Sat Jun 05, 2021 10:41 am "No mean reversion" is the more conservative assumption when it comes to financial planning--you may overestimate future uncertainty, but that's better than underestimating it. It also leads to simpler strategies that don't require market timing. So that's the assumption I favor in my financial planning--it's simple and conservative.
I agree with this.
McQ
Posts: 1425
Joined: Fri Jun 18, 2021 12:21 am
Location: California

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by McQ »

Ben Mathew wrote: Tue May 11, 2021 7:31 pm I use Shiller's dataset to see how a 35/65 portfolio weathered stock market crashes. Shiller's data covers 150 years from 1871 to 2021. I converted Shiller's monthly data to annual data before doing the analysis.

Here's the real growth of $1 in 1871:

Image

Stock is S&P 500 including dividends.
Bond is 10 year US Treasury bond, held for a month and then rebalanced back to 10 year duration every month. Includes interest and price changes.
35/65 portfolio is rebalanced annually.

There are 7 notable stock market crashes over the last 150 years. Cumulative stock market loss

World War I (1916-'20): -41%
Great Depression (1929-'31): -53%
1937 Recession (1937-'41): -39%
Post World War II (1946-'47): -31%
Oil Crisis (1973-'74): -46%
Tech Crash (2000-'02): -39%
Subprime Crisis (2007-'08): -39%

The worst was the Great Depression at -53%.

Cumulative loss of a 35/65 portfolio during these crashes:

World War I (1916-'20): -38%
Great Depression (1929-'31): -4%
1937 Recession (1937-'41): -12%
Post World War II (1946-'47): -25%
Oil Crisis (1973-'74): -25%
Tech Crash (2000-'02): 3%
Subprime Crisis (2007-'08): 1%

Interestingly, the zagging of bonds in a 35/65 portfolio would have erased stock losses during the Great Depression, the Tech Crash and the Subprime Crisis. It would also have reduced the impact of the 1937 recession (-39% reduced to -12%) and the Oil Crisis (-46% reduced to -25%). The worst period for 35/65 was World War I (-38%) when bonds did poorly. Bonds also didn't help much in the post World War II crash (-31% reduced to -25%).

For investors with a balanced 35/65 portfolio, the worst period was not the Great Depression but World War I.

There's also an interesting period from 1977-'81 where the stock market loses only 10%, but 35/65 portfolio does badly (-26%) because of bonds (-35%).

Shillers data is from his spreadsheet "US Stock Markets 1871-Present and CAPE Ratio," available here.
My spreadsheet with this analysis here.
Apologies for posting so late in this thread. It was almost done firing just before I joined BH, and I overlooked it until the namajones post today.

There is a problem with generating returns from the Shiller dataset. Let me explain.

First, there’s no problem with the Shiller stock returns. I beat on the underlying Cowles data, going back to the source in the Commercial & Financial Chronicle, and after months of exacting scrutiny and correction of survivorship bias, the fruit of my labor was barely to lower annualized returns 1871 to 1897 by a dozen basis points. De nada.

But the bond returns posted by Shiller are mis-specified (that’s academese for FUBAR). There was no regular issue of 10-year Treasuries until the 1970s. The bond returns tabled by Shiller are something else entirely. See this thread / post for the messy details: viewtopic.php?p=6131409#p6131409

In short, any bond portfolio that a BH retiree would have bought, had such a one existed in the latter decades of the 19th century, would have returned rather more than what Shiller’s data shows.

Therefore the balanced fund results undershoot what would have been achieved in the years prior to 1914.

Conversely, the OP is absolutely correct about the vicissitudes surrounding WW I. It was a bad, bad time to be an investor, with US inflation briefly worse than around the Civil War, with real returns suffering to match. And if you were a US investor with a global balanced fund, 1910 was about the worst stretch in history to retire and begin making withdrawals (it’s not called the Great War for nothing).

On the other hand, although the Panic of 1873 has an honored place among the crises nommées (along with 1819, 1837, 1857, 1893, 1907, and 1921), on a real total return basis (as opposed to a Main Street suffering basis), it just wasn’t that bad for the buy-and-hold investor. High dividends for reinvestment, plus deflation, kept real returns relatively sweet, especially for an investor who “stayed the course” through 1879, the approximate equivalent then of 1982-83 more recently.
You can take the academic out of the classroom by retirement, but you can't ever take the classroom out of his tone, style, and manner of approach.
smectym
Posts: 1530
Joined: Thu May 26, 2011 5:07 pm

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by smectym »

namajones wrote: Sat Jun 05, 2021 4:55 pm
SquawkIdent wrote: Sat Jun 05, 2021 1:57 pm

As long as it’s good with them, it’s good with me. I paying them to keep me out of trouble. :sharebeer
Isn't Wellesley actively managed? That always makes me nervous. What if the manager changes and the fund goes downhill? Why not just LifeStrategy Conservative Growth Fund? Same mix of stocks and bonds.
In these times especially, you *want* active managing on the bond side, and that’s where the vast majority of Wellesley’s active managing happens.
GaryA505
Posts: 2907
Joined: Wed Feb 08, 2017 1:59 pm
Location: New Mexico

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by GaryA505 »

Some people say 40/60 (or 35/65) is best, but some people say 60/40 (or 65/35) is best. I think I'll split the difference and go with 50/50. :D
Get most of it right and don't make any big mistakes. All else being equal, simpler is better. Simple is as simple does.
seajay
Posts: 1656
Joined: Sat May 01, 2021 3:26 pm
Contact:

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by seajay »

Ben Mathew wrote: Fri Jun 04, 2021 11:51 pm
esteen wrote: Fri Jun 04, 2021 11:36 pm Am I reading that 100% bond line correctly... That from a period of around 1930-1982 (50+ yrs) there was zero gain in purchasing power?
Yes, you're reading that right. $1 invested in 1941 in 10 year Treasury bonds would have been reduced to $0.45 by 1982 after adjusting for inflation.
And even lower after taxation, when for much of that time the lowest rate was 20%

Image

Much of the decline/loss over that period was a result of transitioning from low to high yields. Looking at just the yields only, ignoring price changes, and the net compounded was around 20% lower than the gross

Code: Select all

| Year | 10yr Yield | CPI Index | Lowest Tax | Net Yield | CPI rebased | Gross | Net   | Gross Real | Net real |
|------|------------|-----------|------------|-----------|-------------|-------|-------|------------|----------|
| 1940 |            | 13.9      |            |           | 1           | 1     | 1     | 1          | 1        |
| 1941 | 1.95       | 14.1      | 10         | 1.76      | 1.014       | 1.020 | 1.018 | 1.005      | 1.003    |
| 1942 | 2.46       | 15.7      | 19         | 1.99      | 1.129       | 1.045 | 1.038 | 0.925      | 0.919    |
| 1943 | 2.47       | 16.9      | 19         | 2.00      | 1.216       | 1.070 | 1.059 | 0.880      | 0.871    |
| 1944 | 2.48       | 17.4      | 23         | 1.91      | 1.252       | 1.097 | 1.079 | 0.876      | 0.862    |
| 1945 | 2.37       | 17.8      | 23         | 1.82      | 1.281       | 1.123 | 1.098 | 0.877      | 0.858    |
| 1946 | 2.19       | 18.2      | 20         | 1.75      | 1.309       | 1.148 | 1.118 | 0.876      | 0.854    |
| 1947 | 2.25       | 21.5      | 20         | 1.80      | 1.547       | 1.173 | 1.138 | 0.759      | 0.736    |
| 1948 | 2.44       | 23.7      | 20         | 1.95      | 1.705       | 1.202 | 1.160 | 0.705      | 0.680    |
| 1949 | 2.31       | 24        | 20         | 1.85      | 1.727       | 1.230 | 1.182 | 0.712      | 0.684    |
| 1950 | 2.32       | 23.5      | 20         | 1.86      | 1.691       | 1.258 | 1.203 | 0.744      | 0.712    |
| 1951 | 2.57       | 25.4      | 20.4       | 2.05      | 1.827       | 1.291 | 1.228 | 0.706      | 0.672    |
| 1952 | 2.68       | 26.5      | 22.2       | 2.09      | 1.906       | 1.325 | 1.254 | 0.695      | 0.658    |
| 1953 | 2.83       | 26.6      | 22.2       | 2.20      | 1.914       | 1.363 | 1.281 | 0.712      | 0.670    |
| 1954 | 2.48       | 26.9      | 20         | 1.98      | 1.935       | 1.396 | 1.307 | 0.722      | 0.675    |
| 1955 | 2.61       | 26.7      | 20         | 2.09      | 1.921       | 1.433 | 1.334 | 0.746      | 0.694    |
| 1956 | 2.9        | 26.8      | 20         | 2.32      | 1.928       | 1.474 | 1.365 | 0.765      | 0.708    |
| 1957 | 3.46       | 27.6      | 20         | 2.77      | 1.986       | 1.526 | 1.403 | 0.768      | 0.706    |
| 1958 | 3.09       | 28.6      | 20         | 2.47      | 2.058       | 1.573 | 1.437 | 0.764      | 0.699    |
| 1959 | 4.02       | 29        | 20         | 3.22      | 2.086       | 1.636 | 1.484 | 0.784      | 0.711    |
| 1960 | 4.72       | 29.3      | 20         | 3.78      | 2.108       | 1.713 | 1.540 | 0.813      | 0.730    |
| 1961 | 3.84       | 29.8      | 20         | 3.07      | 2.144       | 1.779 | 1.587 | 0.830      | 0.740    |
| 1962 | 4.08       | 30        | 20         | 3.26      | 2.158       | 1.851 | 1.639 | 0.858      | 0.759    |
| 1963 | 3.83       | 30.4      | 20         | 3.06      | 2.187       | 1.922 | 1.689 | 0.879      | 0.772    |
| 1964 | 4.17       | 30.9      | 16         | 3.50      | 2.223       | 2.003 | 1.748 | 0.901      | 0.786    |
| 1965 | 4.19       | 31.2      | 14         | 3.60      | 2.245       | 2.086 | 1.811 | 0.930      | 0.807    |
| 1966 | 4.61       | 31.8      | 14         | 3.96      | 2.288       | 2.183 | 1.883 | 0.954      | 0.823    |
| 1967 | 4.58       | 32.9      | 14         | 3.94      | 2.367       | 2.283 | 1.957 | 0.964      | 0.827    |
| 1968 | 5.53       | 34.1      | 14         | 4.76      | 2.453       | 2.409 | 2.050 | 0.982      | 0.836    |
| 1969 | 6.04       | 35.6      | 14         | 5.19      | 2.561       | 2.554 | 2.157 | 0.997      | 0.842    |
| 1970 | 7.79       | 37.8      | 14         | 6.70      | 2.719       | 2.753 | 2.301 | 1.012      | 0.846    |
| 1971 | 6.24       | 39.8      | 14         | 5.37      | 2.863       | 2.925 | 2.425 | 1.022      | 0.847    |
| 1972 | 5.95       | 41.1      | 14         | 5.12      | 2.957       | 3.099 | 2.549 | 1.048      | 0.862    |
| 1973 | 6.46       | 42.6      | 14         | 5.56      | 3.065       | 3.299 | 2.690 | 1.077      | 0.878    |
| 1974 | 6.99       | 46.6      | 14         | 6.01      | 3.353       | 3.530 | 2.852 | 1.053      | 0.851    |
| 1975 | 7.5        | 52.1      | 14         | 6.45      | 3.748       | 3.795 | 3.036 | 1.012      | 0.810    |
| 1976 | 7.74       | 55.6      | 14         | 6.66      | 4.000       | 4.088 | 3.238 | 1.022      | 0.809    |
| 1977 | 7.21       | 58.5      | 0          | 7.21      | 4.209       | 4.383 | 3.471 | 1.041      | 0.825    |
| 1978 | 7.96       | 62.5      | 0          | 7.96      | 4.496       | 4.732 | 3.748 | 1.052      | 0.833    |
| 1979 | 9.1        | 68.3      | 0          | 9.10      | 4.914       | 5.163 | 4.089 | 1.051      | 0.832    |
| 1980 | 10.8       | 77.8      | 0          | 10.80     | 5.597       | 5.720 | 4.530 | 1.022      | 0.809    |
| 1981 | 12.57      | 87        | 0          | 12.57     | 6.259       | 6.439 | 5.100 | 1.029      | 0.815    |
| 1982 | 14.59      | 94.3      | 0          | 14.59     | 6.784       | 7.379 | 5.844 | 1.088      | 0.861    |
seajay
Posts: 1656
Joined: Sat May 01, 2021 3:26 pm
Contact:

Re: 35/65 portfolio during stock market crashes, 1871-2021

Post by seajay »

GaryA505 wrote: Tue Aug 24, 2021 12:35 am Some people say 40/60 (or 35/65) is best, but some people say 60/40 (or 65/35) is best. I think I'll split the difference and go with 50/50. :D
80/20 stock/corporate bonds, or 90/10 stock/T-Bills

$40Tn stock market cap, $10Tn Corporate bond market cap. By buying the corporate bonds that stocks issue you eliminate leverage to leave just stock exposure. Leverage just broadly scales volatility, has no effect on rewards. So 80/20 has the same broad reward expectancy, less volatility. The higher volatility (all-stock) just zigzags around the lower volatility (80/20). Corporate bonds reward more than Treasury, but have higher default rate, washes. Shift bond risk over to the stock side a.k.a Buffett 90/10 stock/T-Bills .. and again the same.

Outside of that hold enough bonds to liability match expected/potential spending. If 2%/year to supplement pensions is 'enough', 30 years expectancy, 60% bonds, rest in 'stock' but where 'stock' might be 90/10 stock/T-Bills. 36/60/4 stock/bonds/T-Bills. That as bonds are spent transitions to 90/10 stock/T-Bills, averages 63/30/7. Rounding to 60/40 or 67/33 ..etc. constant weighted likely sees little overall differences.

If valuations are high at the start, buckets can be better, transition from low stock to higher stock over a period when stocks might dip/dive. If valuations are low then constant weighted (60/40 or whatever) is likely better. Present low interest rate era and shortening down on bonds (closer to maturity) and buckets rather than constant weight is potentially the better choice. But that's timing, so as you say split the difference and you'll be neither fully right nor fully wrong.
Post Reply