What's up with bonds?

[nomadscientist]

100% bond portfolio, 4% WR, 30 year lifetime, 0% fees

FIRECalc - "long interest":


51% failure rate

---

cFIREsim:


"half-ish"? failure rate

---

Big ERN's 10yr treasury total return (my re-implementation):


53% failure rate

----

Trinity Study 100% bond real terms:

https://www.aaii.com/files/pdf/6794_ret ... inable.pdf

80% failure rate

----

Any idea what is going on here? There is a substantial divergence between the online calcs and the Trinity Study.

[nomadscientist]

For those who already read this thread - there was an error in my plotting tool that was cutting off some of the better performing scenarios for bonds with Big ERN's data (I think this is a combo of unexpected deflation and decreasing yields). So, now all the online calculators are more or less consistent.

There is still a big disagreement with the Trinity Study, though.

[classical_Liberal]

I'm guessing this is the problem:

The Standard & Poor’s 500 index was used to represent stocks, and long-term, high-grade corporate bonds were used to represent bonds. (All stock, bond, and inflation data were from “Stocks, Bonds, Bills, and Inflation, 1996 Yearbook,” Ibbotson Associates, 1996

The data sets for bonds were different. @Tyler9000 is the historical data finding expert to ask, but the "high grade corporate bonds" used by the cited study are likely much different data than being used by the online calculators. Particularly those using US treasuries as the bond. Plus, there's likely more scenarios being run by the online calculators using data after 1996, which has mostly been a cyclical bull bond market. See all those big winners from 1975 on in the cfiresim image? Those weren't included in the cited study because they didn't have 30 years of data yet.

Honestly, I never buy any corporate bonds as they tend to work counter to my noncoorlation strategy. Cfiresim is basically useless for AA data analysis unless you are using some derivation of a total US stock/total US bond bogleheads's portfolio.

[classical_Liberal]

I thought about this a bit more. I mean in general, the idea of a bond only portfolio. To attempt something like this I would investigate a portfolio built of: 25% corporate bonds, probably junk (for the yield). 25% LT Treasuries (yield/rate sensitivity) 25% tbills (for deflation/low rate senstivity) and 25% I-term TIPS (to cover inflation).


I would play with that a bit to see if it's viable at any reasonable WR for any reasonable period. Honestly though, I think having a 25% stock component would outperform the corporate bonds, but if we are just playing theory with bond only, the above allocation would be my starting point.

[Lucky C]

Besides comparing corporate bonds to government bonds and perhaps more importantly, you're ignoring the impact of the years involved in the Trinity Study, 1926-1995, vs the broader firecalc/cfiresim set.

Set cfiresim to 1926-1995, 100% bonds, and you can even set fees down to 0.05% to see that this is a terrible span of time to have a high bond allocation. Or just look at the purple to blue lines (starting 1926 to 1965) in your plot above. It starts out doing well for a few years when you would prefer to be in bonds vs. stocks during the Great Depression, but 10-year Treasury rates are already < 4% to start this period and all they have going for them is that they're in a bull run that bottoms in 1941 at 10-year rates below 2%. Then treasuries begin their secular bear run from the 1941 (< 2%) all the way to to 1982 (> 15%).

The current (ending?) 38-year bond secular bull run started in 1982, so that is when you start to see good returns from portfolios heavy in bonds for the first time since the Great Depression. The Trinity Study would only include 1965-1995 for its final 30-year period, so these recent bond bull runs are not included. Prior to 1982-2012, the previous 30 year period that would be fully within a secular bond bull market would be 1911-1941, which would be too early for the Trinity Study.

The only portfolios that don't fail cfiresim that fall within the Trinity Study years (100% bonds with fees set to 0.05%) are 1926-1932, 1952, and 1953. You're left with 2.4% at the end of the 1952 starting run and 1.2% in 1953, so 1926-1932 were the only years that your portfolio would last 31+ years. If you count 1952 & 1953 as successes then that's 9 successful years out of the 41 periods of 30 years stated by the Trinity Study, or 78% failure rate. If you want your portfolio to last 31+ years there are 7 of 41 successes or 83% failure rate. This matches the Trinity Study 80% figure even though they used corporate bonds rather than treasuries.

So I think that answers the question about the data discrepancy. Another important question is: Now that 10-year rates are below 0.7%, which is well below the lowest rate in the historical data (1.95% in 1941), and we are on the 38th year of the secular bond bull run which follows a 41-year secular bond bear run, do you really want to base any decisions about what bond allocation you should use for the next 30 years on any of these sets of data?

[classical_Liberal]

Another important question is: Now that 10-year rates are below 0.7%, which is well below the lowest rate in the historical data

Right! I probably should have mentioned this too, if OP is actually thinking about a total bond portfolio. I was mostly just thinking in terms of AA theory, not actual current investments. Now is probably the worst possible time to look at actually trying an all bond portfolio with real money.

[Lucky C]

Maybe post-crisis there will be a country or set of countries in the developing world with double-digit bond rates that have the potential to go down over several decades, the equivalent of where the US was in the early 80's. If their currency also has the potential to gain strength vs. USD then that could be a great option for fixed income diversification. Probably too early to jump in now, but if some foreign bond markets can potentially enter decades-long bull markets then it's OK to be a year or two late. With the demand for any decent yield these days there probably isn't a good option now anyway, but worth keeping an eye on if rates spike back up.