Stocks for the Long Run? Sometimes yes. Sometimes no.

[McQ]

I’m pleased to announce that my re-examination and reconstruction of historical US stock and bond returns has now been published in the Financial Analysts Journal: https://www.tandfonline.com/doi/full/10 ... 23.2268556.

If the FAJ link hits a paywall, I’ve left the underlying working paper up at SSRN: https://papers.ssrn.com/sol3/papers.cfm ... id=3805927

I’ve retitled the working paper as “Where Siegel Went Awry: Outdated Sources and Incomplete Data.” That’s not a bad summary of why I get different results for the 19th century compared to what Jeremy Siegel reports in his book, Stocks for the Long Run. I also have access to much more international history than Siegel had when he first advanced his thesis in 1992.

Briefly on the findings:

1. Stocks have not always beat bonds over multi-decade holding periods here in the US. Recency bias—returns following WW II—have heavily shaped the conventional wisdom.

2. Internationally, one can show stocks losing money in real terms even when held over arbitrarily long intervals of 20, 30, and 50 years. And that’s after excluding nations that lost a war, that suffered civil war, or got invaded and occupied during war. I base the statement on intact nations and periods.

3. Neither the correlation between stocks and bonds nor the relative size of their standard deviations has been stable over time.

4. The contribution of dividends to total return has also not been stable. There were multi-decade periods before 1900 where price appreciation was nil or negative, and all the total return came from dividends.

The 20th century US, as viewed through the lens of the Stocks, Bonds, Bills & Inflation yearbook, emerges as rather an outlier. Results do not generalize to other times and places.

Happy to discuss here on the forum.

[nisiprius]

I'm delighted to see this. You comment in the abstract, "Contrary to Siegel, the pattern of asset returns seen in the 20th century does not generalize to the 19th century,"

This reminded me of William J. Bernstein's essay, Only Two Centuries of Data. Based on Siegel's data, he points out that (even within that data), during 1901-2000, stocks beat bonds by almost 5% annualized... but during 1801-1900, the difference was only about 1.5%.

If we believe the 5% margin of stocks over bonds in the 20th century, then stocks are a nearly riskless investment in the long term... but if the 19th century data are relevant, then stocks clearly are riskier than bonds.

He theorizes that

With hindsight we can see that the 5% stock-bond return gap in the 20th century was the result of a totally unexpected inflationary burst produced by the abandonment of hard money. You can’t abandon hard money twice, so a repeat is not possible.

[Tib]

Does the study identify conditions that seem to explain stock-outperformance during certain eras? If so, do those stock-favoring conditions exist now and look likely to continue to exist for a while?

[Broken Man 1999]

Seems one should own stocks AND bonds, no?

Broken Man 1999

[sc9182]

Money is literally created out of thin air - where were (and what valuation) the magnificent-7 or FAANG were at - a decade or two ago !? Yes, there is some cost to produce their products and services - but we know how profitable (and valuable) those companies are!

If you held those company stocks individually or thru an index — you know the value those companies added to the market/portfolio.

So, hard-money aspects of valuation models don’t necessarily apply any-longer.

Today - every family has 3-4 phones, most likely, smart-phones at that. Wants become needs, and civilization progresses (along with increased productivity gains).

So long as - investors are allowed to participate in par-taking in ownership of the companies (thru “stock” ownership) - we, as investors bound to gain/rewarded.

Yes - companies need bonds, banks (loans, lines of credits among other things), and Wall-Street for their funding/operation. And -
those are already in-place ..

Investors ReJoice !

[er999]

If someone believes stocks will have a terrible return over 30-40 years, likely that person will be better off going for a government job with a pension rather than a bond heavy portfolio beginning in their 20s.

Still a good reminder that once you’ve accumulated sufficient assets, take risk off the table with bonds (preferably tips as I know you are a fan of per your other threads)

[HeavyChevy]

I'm not sure why tracking stocks over a very long period is any more enlightening than tracking horse population, train miles travelled, electricity use, nuclear power use, solar energy use, computer power, etc.

The world changes and conditions even 100 years ago are greatly dissimilar from conditions now.

[Taylor Larimore]

McQ:

Nice job!

Thank you.

Taylor

Jack Bogle's Words of Wisdom: "The Bogleheads forum is a treasure trove of information, as its contributors and visitors alike help one another, with no axe to grind, in all matter of topics that essentially cover the entire field of investing."

[spencydub]

If someone believes stocks will have a terrible return over 30-40 years, likely that person will be better off going for a government job with a pension rather than a bond heavy portfolio beginning in their 20s.

Still a good reminder that once you’ve accumulated sufficient assets, take risk off the table with bonds (preferably tips as I know you are a fan of per your other threads)

What was his take on TIPS? And do iBonds fit the bill?

[TipsQuestions]

Nice work sir. I've wasted enough time studying historical data to conclude that reconstituted stock returns from decades or centuries ago should be given little weight for current financial planning. The world was too different, and the found data is too sketchy. Even DFA will tell you that their pre 1962 (Fama/French) data was gathered by hand and is less accurate than data from after. This is especially true if old data results in something irrational, like an ERP of 0 over a long period of time.

[McQ]

I'm delighted to see this. You comment in the abstract, "Contrary to Siegel, the pattern of asset returns seen in the 20th century does not generalize to the 19th century,"

This reminded me of William J. Bernstein's essay, Only Two Centuries of Data. Based on Siegel's data, he points out that (even within that data), during 1901-2000, stocks beat bonds by almost 5% annualized... but during 1801-1900, the difference was only about 1.5%.

If we believe the 5% margin of stocks over bonds in the 20th century, then stocks are a nearly riskless investment in the long term... but if the 19th century data are relevant, then stocks clearly are riskier than bonds.

He theorizes that

With hindsight we can see that the 5% stock-bond return gap in the 20th century was the result of a totally unexpected inflationary burst produced by the abandonment of hard money. You can’t abandon hard money twice, so a repeat is not possible.

Thanks nisiprius, good to have this quote from Bill (I cited a different paper of his). The quotes you pulled nicely gloss the theme of my paper: non-stationarity. Regimes change.

And I have come around to the idea that the demise of hard currency in the 1930s was one of the biggest changes in regime.

PS: at the very end of the FAJ paper, in the Conclusions section, I work in Mandelbrot and Hudson--not least because of your comments here in the forum.

[McQ]

Does the study identify conditions that seem to explain stock-outperformance during certain eras? If so, do those stock-favoring conditions exist now and look likely to continue to exist for a while?

Nope. It is entirely concerned with accurate retrodiction. Explanations, if any, are left to future research, see the conclusions section at the end of the main text.

[McQ]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

[HomerJ]

I don't care about 1800-1900...

You might make a case that 1900-2000 is no longer as relevant, but you can never convince me that 1800-1900 IS relevant...

But, of course, I could be wrong... but I will likely never know...

[HomerJ]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

Heh, you answered my post before I wrote my post.

You don't sound very serious here, but I'll give you a serious answer.

Anything before the Fed was established (1913) or maybe the Securities Acts of 1933 and 1934, I don't care about.

I'm curious how reliable your DATA is from the 1700s and 1800s, but even if it is reliable, I don't find it relevant.

My apologies. No one wants to hear that their hard work isn't relevant to the modern world. Still interesting from a historical perspective.

Again maybe I'm wrong. Note I'm 50/50 stocks/bonds so I'm probably more in agreement with you than not. But I don't think anything from the 1800s is relevant.

[Ocean77]

US. Recency bias—returns following WW II

I had not thought of the years or even decades post WWII as "recent".

[Thesaints]

Thank you for posting your article.
There are a couple of minor points, not extremely consequential per se, but they hint to what the philosophy of the article is as a whole.
- On Table 1 you report that the data set covers 226 single years, 222 5-year periods, 217 10-year periods, and so on.
That is true in a way, but for multiple year periods those periods are not independent from each other.
In fact, one could very well argue that the 50-year period from 1950 to 1999 and the period 1951-2000 are pretty much one and the same and you are double counting them. Over 200 years there are only four 50-year periods truly different from each other and of course there are a few more quite different from each other and many more almost identical to each other.
- My second point is when in the same table you write "odds that stocks beat bonds". Those are not odds, that is past data.
if we just stop at the 50-year period analysis, what you are showing is that earlier on bonds were better, then stocks and bonds were about the same, and finally stocks have been better (and this because of the non-independence of consecutive periods I mentioned earlier).
That's what happened; we do not know the odds of that happening.

Overall, you make a rather exhaustive analysis of historical data, cutting and dicing it every which way and that is valuable. What is missing is an hypothesis of a mechanism behind the fact that stocks are better (or that bonds are, or that the two are about equivalent).
But that mechanism is known and it points to stocks having a higher expected return. That is because bonds return is contractually set and it is quite decoupled from the state of the underlying business insofar debt is not defaulted.
That is the reason why investment grade bonds are little correlated to stocks, while junk bonds have a much higher correlation.
For the former, probability of default is very small and it does not matter if the issuer does very well, so-so, or badly: the bond holder will collect the agreed upon interest in almost all cases. Nothing more, nothing less.
The situation is different for junk bonds where there is a higher chance of default and issuers not doing very well will much increase their default rate, while issuers doing well will much decrease it, and default there is also a big factor in realized returns, in addition to the nominal yield.

On the other hand, the return from stocks does depend primarily on business results (although in a relative fashion, the stock of a company "doing well" in absolute terms may not do nearly as well if other companies do much better at the same time).
And that's all we need to know: if the two classes had the same expected return, but the return from bonds is much less uncertain, who would invest in stocks ? At most you would have few investors investing in stocks in the short term, when they need the chance of an extra return, but that chance would go to zero in the long term, so nobody in their sane mind would invest in stocks for the long term.
But investors do and, in most cases, they are not insane.

[zero_coupon]

I’m pleased to announce that my re-examination and reconstruction of historical US stock and bond returns has now been published in the Financial Analysts Journal: https://www.tandfonline.com/doi/full/10 ... 23.2268556.

Excellent news. I referred someone to this paper just this past Monday! See here:

I would recommend reading Dr. Bernstein's "Four Pillars of Investing" that details the history of the performance of various asset classes over centuries.

Also see McQ's paper on this. Goes further back in time.

[Svensk Anga]

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

I think the relevance line belongs at the end of the gold standard. So 1933. Since that time, nominal bonds have no longer been as reliable a store of value as on the gold standard. It became better to hold real assets, represented by equity. Maybe the 22nd century McQ will draw another line at 1997, the introduction of TIPS.

At about the same time (1934), the SEC was established. Prior to this, equity investor protections under the law were minimal and the theoretical pre-Depression Boglehead was probably well advised to stay away from stocks much as present day Bogleheads are advised to stay away from crypto.

[Calvert]

“Where Siegel Went Awry: Outdated Sources and Incomplete Data.” That’s not a bad summary of why I get different results for the 19th century compared to what Jeremy Siegel reports in his book, Stocks for the Long Run. I also have access to much more international history than Siegel had when he first advanced his thesis in 1992.

McQ ... I commend you for your data gathering and measurement efforts ! I'm a bit jealous of the nuances you must have seen ... that can't fit into a paper (or even a book).

Sadly, I very much enjoyed "Stocks for the Long Run". I recently bought and read the 6th addition.

I wanted to share a quoted exchange from another book I've enjoyed. I've edited it for clarity.

"When did Lameth write his book?"
"Oh - I should say about eight hundred years ago. Of course he has based it largely on the previous work of Gleen."

"Then why rely on him? Why not go to ... and study the remains for yourself?"
Dorwin raised his eyebrows ... "Why whatever for?"

"To get the information firsthand, of course."

"But where's the necessity? It seems an uncommonly round about and hopelessly rigmarolish method of getting anywhere. Look here, now, I've got the works of all the old masters - the great ... of the past. I weigh them against each other - balance the disagreements - analyze the conflicting statements - decide which is probably correct - and come to a conclusion. That is the scientific method. ... How insufferably crude it would be to go there and bunder about, when the old masters have covered the ground so much more effectually than we could possibly hope to do."

[GAAP]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

To me, it depends upon what you are trying to achieve. Historical results are useful for evaluating how a strategy might react under different conditions. I find them less useful for developing a strategy.

As for relevance, the end of the gold standard, legal gold ownership, ERISA, and TIPS led to fairly major changes in the environment. Relevance depends upon what you are attempting to do with the data.

[petulant]

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

I think the relevance line belongs at the end of the gold standard. So 1933. Since that time, nominal bonds have no longer been as reliable a store of value as on the gold standard. It became better to hold real assets, represented by equity. Maybe the 22nd century McQ will draw another line at 1997, the introduction of TIPS.

At about the same time (1934), the SEC was established. Prior to this, equity investor protections under the law were minimal and the theoretical pre-Depression Boglehead was probably well advised to stay away from stocks much as present day Bogleheads are advised to stay away from crypto.

Problem is, that wasn't the end of the gold standard. The U.S. continued to operate a gold standard and exchanged gold with other countries' financial institutions until the 1970s. Only at the time was the link with gold fully and finally severed. The terrible performance of bonds 1945-1981 has something to do with the shifts in the monetary regime.

[abc132]

I commend the quest for better data.

I don't think it is useful to measure periods of maximum underperformance for stocks and bonds. Almost none of us have a set amount of money without accumulation or decumulation and our results will always be much different than the worst-case presentation of a lumpsum invested in a single year at a single worst moment in time.

We now have dueling threads for people to use to be scared of bonds and to be scared of stocks. This methodology of "worst period" is definitely not actionable as it is typically used. It is not something to try and avoid because of loss aversion.

I recommend the following questions for stocks
1) Do you think companies are going to be profitable and generate profits over your investing timeline?
2) Are company profits going to be shared with investors through stock prices?

For stocks I think the answer to both today is YES and there is a strong case for stock ownership. In the US we see 200x the average wage from company CEO's and we see much of their compensation based on stock prices, despite the presence of golden parachutes.

It's not so much that regimes change but that our choices as investors have changed. In 30 seconds and for zero purchase fees a middle class person can participate in the stock market. This is so different from the past, in terms of who can do this, how easy it is, and how low cost it is. Having a voting population that is invested in getting a shares of those stock returns means that is is tougher (but definitely not impossible!) for businesses as a whole to extract the value from investors.

Stocks are here for the long run - with the long run being our lifetimes.

[Call_Me_Op]

I think we need to be careful here, because our friend McQ is talking about long (nominal) bonds. Few here hold them in any significant allocation, and we know that they are about as risky as stocks - so not surprising there can be periods where they have higher return.

[StillGoing]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

McQ congratulations on the paper.

My position is that historical data offers useful guidance particularly, as others have said, for comparing different accumulation or decumulation strategies.

The question I might ask is "could the observed historical returns, particularly periods of poor performance, reoccur in a similar fashion in the future?"

For example,
1) Could high inflation affect real returns over an extended period of time (i.e., a rerun of the 1970s)?
2) Could the stock market crash by more than 50% over the course of a year or two (i.e., a rerun of 1929)?
3) Could real bond returns be low or negative over periods of a decade or more (a rerun of the 1940s onwards)?

My answer to each of these questions is 'yes', and this feeds in to my using historical returns as an indicative, but not predictive, tool. But it also argues for a more sophisticated analysis of such results than looking at a single number (e.g., the oft quoted 4% safe withdrawal rate) since 'optimising' for the retirement that provides the 'safe withdrawal rate' may have significant knock on effects on other retirements.

I'm currently working with a data set that goes back to the 17th century (UK government bonds) but have drawn a line for calculating returns at 1870 since this is the earliest date that UK treasury bills were issued (although from 1870 to about 1918, bankers bills predominated in terms of the amounts in issue, so I've used those instead) and, as far as I am aware, there is no freely available total return series for the UK stock market prior to that date. An argument could also be made for drawing the line at 1900 when the first dated gilt was issued (only undated gilts, or consols, were available before that) or 1915 when dated gilts were issued in reasonable quantities. So, practical issues may be more important.

cheers
StillGoing

[McQ]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

Heh, you answered my post before I wrote my post.

You don't sound very serious here, but I'll give you a serious answer.

Anything before the Fed was established (1913) or maybe the Securities Acts of 1933 and 1934, I don't care about.

I'm curious how reliable your DATA is from the 1700s and 1800s, but even if it is reliable, I don't find it relevant.

My apologies. No one wants to hear that their hard work isn't relevant to the modern world. Still interesting from a historical perspective.

Again maybe I'm wrong. Note I'm 50/50 stocks/bonds so I'm probably more in agreement with you than not. But I don't think anything from the 1800s is relevant.

I t was a serious question, philosophical even. Your two proffered demarcation lines, 1913 and 1934, are quite reasonable, if data must be discarded. Certainly more reasonable than 1871 and 1926, arbitrary endpoints for data collection.

I personally prefer to know the entire data record, with no omissions.

[McQ]

Thank you for posting your article.
... What is missing is an hypothesis of a mechanism behind the fact that stocks are better (or that bonds are, or that the two are about equivalent).
But that mechanism is known and it points to stocks having a higher expected return. That is because bonds return is contractually set and it is quite decoupled from the state of the underlying business insofar debt is not defaulted.
That is the reason why investment grade bonds are little correlated to stocks, while junk bonds have a much higher correlation.
For the former, probability of default is very small and it does not matter if the issuer does very well, so-so, or badly: the bond holder will collect the agreed upon interest in almost all cases. Nothing more, nothing less.
The situation is different for junk bonds where there is a higher chance of default and issuers not doing very well will much increase their default rate, while issuers doing well will much decrease it, and default there is also a big factor in realized returns, in addition to the nominal yield.

On the other hand, the return from stocks does depend primarily on business results (although in a relative fashion, the stock of a company "doing well" in absolute terms may not do nearly as well if other companies do much better at the same time).
And that's all we need to know: if the two classes had the same expected return, but the return from bonds is much less uncertain, who would invest in stocks ? At most you would have few investors investing in stocks in the short term, when they need the chance of an extra return, but that chance would go to zero in the long term, so nobody in their sane mind would invest in stocks for the long term.
But investors do and, in most cases, they are not insane.

I'm familiar with the theory: stocks are risky, therefore investors demand a higher return to own stocks. Which means that stocks should deliver that higher return over the long term.

Pity the theory didn't survive the confrontation with data.

Alternatively, the theory, naively put, sets up a conundrum. Here is a quote from the Conclusions section:

"The new historical record addresses a conundrum that gradually emerged as the work of Roger Ibbotson and Jeremy Siegel diffused to a broader public. If stocks are risky, investors will demand a premium to invest. But if stocks cease to be risky once held for a long enough period—if stocks are certain to have strong returns after twenty years, and certain to outperform bonds—then investors have no reason to expect a premium over these longer periods, given that no shortfall risk had to be assumed. The expanded historical record shows that stocks can perform poorly in absolute terms, and underperform bonds, whether the holding period is twenty, thirty, fifty, or 100 years. That documentation of risk resolves the conundrum. There can be no equity premium unless stocks are risky; but with that risk documented—even if infrequent in occurrence—investors can, based on the entirety of the historical record, probabilistically expect stocks to outperform bonds over other intervals, including the interval that corresponds to their own investment horizon. Conversely, if stocks had never done poorly outside of war and disaster, there would be no reason to expect stocks to continue to do well in peacetime, because no shortfall risk would have been demonstrated. The fact of a persistent equity premium, and not just its level, would be the puzzle (Mehra and Prescott 1985)."

[McQ]

“Where Siegel Went Awry: Outdated Sources and Incomplete Data.” That’s not a bad summary of why I get different results for the 19th century compared to what Jeremy Siegel reports in his book, Stocks for the Long Run. I also have access to much more international history than Siegel had when he first advanced his thesis in 1992.

McQ ... I commend you for your data gathering and measurement efforts ! I'm a bit jealous of the nuances you must have seen ... that can't fit into a paper (or even a book).

Sadly, I very much enjoyed "Stocks for the Long Run". I recently bought and read the 6th addition.

I wanted to share a quoted exchange from another book I've enjoyed. I've edited it for clarity.

"When did Lameth write his book?"
"Oh - I should say about eight hundred years ago. Of course he has based it largely on the previous work of Gleen."

"Then why rely on him? Why not go to ... and study the remains for yourself?"
Dorwin raised his eyebrows ... "Why whatever for?"

"To get the information firsthand, of course."

"But where's the necessity? It seems an uncommonly round about and hopelessly rigmarolish method of getting anywhere. Look here, now, I've got the works of all the old masters - the great ... of the past. I weigh them against each other - balance the disagreements - analyze the conflicting statements - decide which is probably correct - and come to a conclusion. That is the scientific method. ... How insufferably crude it would be to go there and bunder about, when the old masters have covered the ground so much more effectually than we could possibly hope to do."

Thanks for those kind words (and the book quote ).

Re the nuances that had to be omitted: take a look at this preparatory paper, the last one I wrote before the published sum up: https://papers.ssrn.com/sol3/papers.cfm ... id=3740190

I stuffed pretty much everything left over / didn't fit into the Appendices. There's a list upfront to help you browse.

[McQ]

I think we need to be careful here, because our friend McQ is talking about long (nominal) bonds. Few here hold them in any significant allocation, and we know that they are about as risky as stocks - so not surprising there can be periods where they have higher return.

Absolutely correct. In the first 40 years of the sample US Treasuries had no maturity date. Many of the municipals through the 1890s also had no maturity date. The corporate bonds from 1897 to 1926 included maturities as long as 100 years.

After 1926, when I revert to existing sources (the SBBI) it is a 20-year bond.

[McQ]

...

McQ congratulations on the paper.

My position is that historical data offers useful guidance particularly, as others have said, for comparing different accumulation or decumulation strategies.

The question I might ask is "could the observed historical returns, particularly periods of poor performance, reoccur in a similar fashion in the future?"

For example,
1) Could high inflation affect real returns over an extended period of time (i.e., a rerun of the 1970s)?
2) Could the stock market crash by more than 50% over the course of a year or two (i.e., a rerun of 1929)?
3) Could real bond returns be low or negative over periods of a decade or more (a rerun of the 1940s onwards)?

My answer to each of these questions is 'yes', and this feeds in to my using historical returns as an indicative, but not predictive, tool. But it also argues for a more sophisticated analysis of such results than looking at a single number (e.g., the oft quoted 4% safe withdrawal rate) since 'optimising' for the retirement that provides the 'safe withdrawal rate' may have significant knock on effects on other retirements.

I'm currently working with a data set that goes back to the 17th century (UK government bonds) but have drawn a line for calculating returns at 1870 since this is the earliest date that UK treasury bills were issued (although from 1870 to about 1918, bankers bills predominated in terms of the amounts in issue, so I've used those instead) and, as far as I am aware, there is no freely available total return series for the UK stock market prior to that date. An argument could also be made for drawing the line at 1900 when the first dated gilt was issued (only undated gilts, or consols, were available before that) or 1915 when dated gilts were issued in reasonable quantities. So, practical issues may be more important.

cheers
StillGoing

Those are interesting demarcation lines: first bills, first dated maturity bonds. I would be reluctant to discard the 1753 - 1870 data for Consols, though. Just too interesting to see how the safest bond in the world (at that time) responded to things like the Napoleonic wars, the various panics, etc.

The only pre-1870 UK stock data series I know is from GFD. (but be aware of Campbell, G., Grossman, R.S. and Turner, J.D., 2021. Before the cult of equity: the British stock market, 1829–1929. European Review of Economic History, 25(4), pp.645-679.

Bryan Taylor allows me to share graphs for the stock series. Here's rolling ten year returns on UK stocks and bonds from 1700 (another piece of the initial FAJ submission that fell on the cutting room floor).

Positive bars show stocks outperforming bonds; negative bars show stock under-performance.

[nedsaid]

I'm familiar with the theory: stocks are risky, therefore investors demand a higher return to own stocks. Which means that stocks should deliver that higher return over the long term.

Pity the theory didn't survive the confrontation with data.

Alternatively, the theory, naively put, sets up a conundrum. Here is a quote from the Conclusions section:

"The new historical record addresses a conundrum that gradually emerged as the work of Roger Ibbotson and Jeremy Siegel diffused to a broader public. If stocks are risky, investors will demand a premium to invest. But if stocks cease to be risky once held for a long enough period—if stocks are certain to have strong returns after twenty years, and certain to outperform bonds—then investors have no reason to expect a premium over these longer periods, given that no shortfall risk had to be assumed. The expanded historical record shows that stocks can perform poorly in absolute terms, and underperform bonds, whether the holding period is twenty, thirty, fifty, or 100 years. That documentation of risk resolves the conundrum. There can be no equity premium unless stocks are risky; but with that risk documented—even if infrequent in occurrence—investors can, based on the entirety of the historical record, probabilistically expect stocks to outperform bonds over other intervals, including the interval that corresponds to their own investment horizon. Conversely, if stocks had never done poorly outside of war and disaster, there would be no reason to expect stocks to continue to do well in peacetime, because no shortfall risk would have been demonstrated. The fact of a persistent equity premium, and not just its level, would be the puzzle (Mehra and Prescott 1985)."

In another thread, I teased that Bogleheads could conduct a sit-in at the the New York Stock Exchange holding clever protest signs
and threatening not to leave until our demands for better equity returns were met. We demand better returns!! We demand better returns!! We demand better returns!! There are some of us still around who remember the 1960's. Perhaps someone would volunteer to organize the protest. Unfortunately, Occupy Wall Street is all ready taken, we will have to come up with something different.

The Equity Risk Premium could simply reflect a change in investor perceptions, a change of investor preferences, and a desire to be compensated in capital gains rather than interest; particularly when capital gains income received preferential treatment from the Tax Code. It might be that investors like stocks a whole lot more now than they did back in the 1800's.

I do believe that certain events in history are important: the founding of the Federal Reserve System back in 1913, the two phase withdrawal from the Gold Standard in 1933 and 1973, the passing of the Securities Exchange Act of 1934. These events created big changes in the markets and the economy but that doesn't mean that data before these events is irrelevant. I think data from the 1600's, the 1700's, and the 1800's holds lessons for us today, somehow we imagine that we are so much smarter than our ancestors and I don't believe that to be the case. It just seems like hubris to ignore all pre-Fed, pre-SEC, pre-Whatever data. Things are different now and we are the exception to world financial history. Not!!

Bill Bernstein did say that over centuries and not decades that the returns of stocks and bonds are about the same. McQ's paper greatly expands upon Bernstein's observations.

Who knows? Perhaps the need to raise money in the debt markets will be so great that interest will get preferential tax treatment and not capital gains. Perhaps people will start liking bonds a lot more than they have in the recent past. Perhaps investor perceptions and preferences will shift again. Perhaps bonds will be where the action is.

[nedsaid]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

Heh, you answered my post before I wrote my post.

You don't sound very serious here, but I'll give you a serious answer.

Anything before the Fed was established (1913) or maybe the Securities Acts of 1933 and 1934, I don't care about.

I'm curious how reliable your DATA is from the 1700s and 1800s, but even if it is reliable, I don't find it relevant.

My apologies. No one wants to hear that their hard work isn't relevant to the modern world. Still interesting from a historical perspective.

Again maybe I'm wrong. Note I'm 50/50 stocks/bonds so I'm probably more in agreement with you than not. But I don't think anything from the 1800s is relevant.

I t was a serious question, philosophical even. Your two proffered demarcation lines, 1913 and 1934, are quite reasonable, if data must be discarded. Certainly more reasonable than 1871 and 1926, arbitrary endpoints for data collection.

I personally prefer to know the entire data record, with no omissions.

It is time for Hall of Fame relief pitcher Rollie Fingers to have someone get the 'stache wax out of the locker and wax up that handlebar moustache. Might even be time for a high and tight brushback pitch, keep the batter from crowding the plate. After he brushes off the dust, he might be set up for that wicked slider.

Time for the scowling game face to glare at the batter with, the 'stache provides a good psychological effect, and keep the batter guessing which pitch will be next. Make sure you shake off the catcher at least a couple of times, it communicates that you are calling the pitches and not the catcher. You gotta be in command.

[StillGoing]

...

McQ congratulations on the paper.

My position is that historical data offers useful guidance particularly, as others have said, for comparing different accumulation or decumulation strategies.

The question I might ask is "could the observed historical returns, particularly periods of poor performance, reoccur in a similar fashion in the future?"

For example,
1) Could high inflation affect real returns over an extended period of time (i.e., a rerun of the 1970s)?
2) Could the stock market crash by more than 50% over the course of a year or two (i.e., a rerun of 1929)?
3) Could real bond returns be low or negative over periods of a decade or more (a rerun of the 1940s onwards)?

My answer to each of these questions is 'yes', and this feeds in to my using historical returns as an indicative, but not predictive, tool. But it also argues for a more sophisticated analysis of such results than looking at a single number (e.g., the oft quoted 4% safe withdrawal rate) since 'optimising' for the retirement that provides the 'safe withdrawal rate' may have significant knock on effects on other retirements.

I'm currently working with a data set that goes back to the 17th century (UK government bonds) but have drawn a line for calculating returns at 1870 since this is the earliest date that UK treasury bills were issued (although from 1870 to about 1918, bankers bills predominated in terms of the amounts in issue, so I've used those instead) and, as far as I am aware, there is no freely available total return series for the UK stock market prior to that date. An argument could also be made for drawing the line at 1900 when the first dated gilt was issued (only undated gilts, or consols, were available before that) or 1915 when dated gilts were issued in reasonable quantities. So, practical issues may be more important.

cheers
StillGoing

Those are interesting demarcation lines: first bills, first dated maturity bonds. I would be reluctant to discard the 1753 - 1870 data for Consols, though. Just too interesting to see how the safest bond in the world (at that time) responded to things like the Napoleonic wars, the various panics, etc.

The only pre-1870 UK stock data series I know is from GFD. (but be aware of Campbell, G., Grossman, R.S. and Turner, J.D., 2021. Before the cult of equity: the British stock market, 1829–1929. European Review of Economic History, 25(4), pp.645-679.

Bryan Taylor allows me to share graphs for the stock series. Here's rolling ten year returns on UK stocks and bonds from 1700 (another piece of the initial FAJ submission that fell on the cutting room floor).

Positive bars show stocks outperforming bonds; negative bars show stock under-performance.

Interesting graph - thanks for that.

I guess that I've drawn the demarcation lines there because bills and dated gilts are securities that can still be invested in (the last UK consols were called in 2014/15 - practically, all bar one, were no longer liquid for a few years before that and represented a tiny fraction of outstanding gilts). For the 1870-1900 period, only bills and consols were available so the intermediate maturity indices have been interpolated using a duration matched approach (in other words, the returns still represent those that could have been obtained by contemporaneous investors, but not entirely satisfactorily). I've got two reports underway - one dealing with the construction of the indices and one dealing with the effect of duration on retirement portfolio performance. While I thought I had finalised the former (but possible will reconsider in the light the paper you've linked to), I'd yet to finalise the exact starting point for the analysis of the latter.

I note that for those of us with limited library access, a version (I think) of the Campbell et al. paper is also available at https://papers.ssrn.com/sol3/papers.cfm ... id=3404050

One of the authors had previously pushed the UK stock market indices back to 1870 (Grossman, Richard S. 2002. New Indices of British Equity Prices, 1870–1913. Journal of Economic History 62(1): 121–146.) which form the early returns (before 1907) in the macrohistory database.

I think there has been a fair amount of work on UK consols since, until recently, they represented an unbroken series of returns etc. There's an interesting paper at Ellison, Martin, and Andrew Scott. 2020. "Managing the UK National Debt 1694–2018." American Economic Journal: Macroeconomics, 12 (3): 227-57. DOI: 10.1257/mac.20180263 that looks at various aspects of debt (and, of course, dated gilts too). It is their extended dataset of gilt prices and coupons that I've used to generate sector returns.

cheers
StillGoing

[kd2008]

Most folks don't accumulate substantial amount until they are 35 yrs old or so. A generous life expectancy of 95 yrs would imply 60 yr period as somewhat of a relevant timeframe to study stock and bond returns. Analyses should focus on 30 yr to 60 yr independent periods to ascertain returns. Everything else is blowing gas, and not actionable in my opinion.

[nedsaid]

It is amazing that folks around here regard the Equity Risk Premium just like an immutable law of physics. Indeed, it is the foundation of Boglehead portfolios. If belief in the ERP went away, then I suppose everyone here would be in
50% stock/50% bond portfolios, which isn't the worst idea in the world by the way. That is what John Bogle settled on later in life.

Seems to me that saying that one doesn't care about data before some demarcation line, pre-1913, pre-1934, pre-Whatever just seems like an exercise in denial. Refuse to look at the old data and the cognitive dissonance just goes away. Something in me says that it is a good idea to think the unthinkable sometimes and realize that sometimes there are bad outcomes. There isn't a pot of gold at the end of every rainbow.

I do believe that the Equity Risk Premium will continue and there are logical reasons for believing that. However, I know that this outcome of higher returns for stocks relative to bonds is not guaranteed. As has been stated here eloquently by several posters, if there isn't risk then why should there be a premium?

As I think about it, the idea of a sit-in protest on the floor of the New York Stock Exchange looks better and better. Hold up our clever protest signs, chant our demands for better equity returns, and refuse to leave until our demands are met.
The key is to get press coverage, interviews on live TV, and most important of all get a huge turnout. I am getting so fired up by this, I am going to write my Congressman. By golly, if we demand higher returns then the market will have to give them to us.

But as with all things in these McQ threads, there are more twists and turns than in a Tom Clancy novel. Just as he had me seriously doubting the Equity Risk Premium, then he puts up a chart of UK Stock returns relative to UK Bond returns since 1710 and it screams Equity Risk Premium! It does show a few 20 year periods of bonds outperforming stocks but a simple graph makes the case for the ERP very compelling. A picture is worth 1,000 words.

[CyclingDuo]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

Heh, you answered my post before I wrote my post.

You don't sound very serious here, but I'll give you a serious answer.

Anything before the Fed was established (1913) or maybe the Securities Acts of 1933 and 1934, I don't care about.

I'm curious how reliable your DATA is from the 1700s and 1800s, but even if it is reliable, I don't find it relevant.

My apologies. No one wants to hear that their hard work isn't relevant to the modern world. Still interesting from a historical perspective.

Again maybe I'm wrong. Note I'm 50/50 stocks/bonds so I'm probably more in agreement with you than not. But I don't think anything from the 1800s is relevant.

I t was a serious question, philosophical even. Your two proffered demarcation lines, 1913 and 1934, are quite reasonable, if data must be discarded. Certainly more reasonable than 1871 and 1926, arbitrary endpoints for data collection.

I personally prefer to know the entire data record, with no omissions.

It is time for Hall of Fame relief pitcher Rollie Fingers to have someone get the 'stache wax out of the locker and wax up that handlebar moustache. Might even be time for a high and tight brushback pitch, keep the batter from crowding the plate. After he brushes off the dust, he might be set up for that wicked slider.

Time for the scowling game face to glare at the batter with, the 'stache provides a good psychological effect, and keep the batter guessing which pitch will be next. Make sure you shake off the catcher at least a couple of times, it communicates that you are calling the pitches and not the catcher. You gotta be in command.

The new pitch clock and batter clock doesn't allow enough time for all of that. As a result, thankfully MLB games are now, on average, about 47 minutes shorter than Killers of the Flower Moon. I'm drawing a demarcation line that begins with the 2023 MLB season.

Home runs reversed trend and are back on the rise with all of the new rules changes, and batting averages improved this year. Most excited to see the rule changes altering the stolen base trend of the past 30+ years. The history of professional baseball still shows we might experience various 20-30 year trends along the way, but at least the games are getting shorter.




https://nowtakingthefield.substack.com/ ... le-changes

CyclingDuo

[nedsaid]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

Heh, you answered my post before I wrote my post.

You don't sound very serious here, but I'll give you a serious answer.

Anything before the Fed was established (1913) or maybe the Securities Acts of 1933 and 1934, I don't care about.

I'm curious how reliable your DATA is from the 1700s and 1800s, but even if it is reliable, I don't find it relevant.

My apologies. No one wants to hear that their hard work isn't relevant to the modern world. Still interesting from a historical perspective.

Again maybe I'm wrong. Note I'm 50/50 stocks/bonds so I'm probably more in agreement with you than not. But I don't think anything from the 1800s is relevant.

I t was a serious question, philosophical even. Your two proffered demarcation lines, 1913 and 1934, are quite reasonable, if data must be discarded. Certainly more reasonable than 1871 and 1926, arbitrary endpoints for data collection.

I personally prefer to know the entire data record, with no omissions.

It is time for Hall of Fame relief pitcher Rollie Fingers to have someone get the 'stache wax out of the locker and wax up that handlebar moustache. Might even be time for a high and tight brushback pitch, keep the batter from crowding the plate. After he brushes off the dust, he might be set up for that wicked slider.

Time for the scowling game face to glare at the batter with, the 'stache provides a good psychological effect, and keep the batter guessing which pitch will be next. Make sure you shake off the catcher at least a couple of times, it communicates that you are calling the pitches and not the catcher. You gotta be in command.

The new pitch clock and batter clock doesn't allow enough time for all of that. As a result, thankfully MLB games are now, on average, about 47 minutes shorter than Killers of the Flower Moon. I'm drawing a demarcation line that begins with the 2023 MLB season.

Home runs reversed trend and are back on the rise with all of the new rules changes, and batting averages improved this year. Most excited to see the rule changes altering the stolen base trend of the past 30+ years. The history of professional baseball still shows we might experience various 20-30 year trends along the way, but at least the games are getting shorter.




https://nowtakingthefield.substack.com/ ... le-changes

CyclingDuo

I love it. Twenty year trends in financial markets and twenty year trends in baseball. Demarcation lines drawn with Baseball data and demarcation lines drawn with financial data.

I can see that more than one poster on Bogleheads has 'stache wax in their locker.

Thanks for your post. Masterfully done in the style of Nisiprius.

[Wrench]

Prior to the 1930s, multi-years stretches of deflation were relatively common:
https://www.in2013dollars.com/UK-inflation
The 1800s in particular appears to be a century or so which was generally deflationary.
Since the 30s, there have been few or no years with substantial deflation, much less long stretches of deflation.
It would be interesting to see if the returns of stocks and bonds over 20, 30 or 50 year periods correlate with inflation.

Wrench

[nedsaid]

Prior to the 1930s, multi-years stretches of deflation were relatively common:
https://www.in2013dollars.com/UK-inflation
The 1800s in particular appears to be a century or so which was generally deflationary.
Since the 30s, there have been few or no years with substantial deflation, much less long stretches of deflation.
It would be interesting to see if the returns of stocks and bonds over 20, 30 or 50 year periods correlate with inflation.

Wrench

My best semi-educated guess is that a bit of inflation is good for stocks, particularly Value stocks. Inflation spikes are devastating to both stocks and bonds. Disinflation is good for stocks and bonds but outright deflation is good for quality bonds but can be tough on stocks, a lot depends upon what is causing the deflation. In the latter 1890's, mild deflation wasn't bad because it was driven by productivity gains, particularly in transportation. That is my take anyways.

[Alpha4]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

McQ congratulations on the paper.

My position is that historical data offers useful guidance particularly, as others have said, for comparing different accumulation or decumulation strategies.

The question I might ask is "could the observed historical returns, particularly periods of poor performance, reoccur in a similar fashion in the future?"

For example,
1) Could high inflation affect real returns over an extended period of time (i.e., a rerun of the 1970s)?
2) Could the stock market crash by more than 50% over the course of a year or two (i.e., a rerun of 1929)?
3) Could real bond returns be low or negative over periods of a decade or more (a rerun of the 1940s onwards)?

My answer to each of these questions is 'yes', and this feeds in to my using historical returns as an indicative, but not predictive, tool. But it also argues for a more sophisticated analysis of such results than looking at a single number (e.g., the oft quoted 4% safe withdrawal rate) since 'optimising' for the retirement that provides the 'safe withdrawal rate' may have significant knock on effects on other retirements.

I'm currently working with a data set that goes back to the 17th century (UK government bonds) but have drawn a line for calculating returns at 1870 since this is the earliest date that UK treasury bills were issued (although from 1870 to about 1918, bankers bills predominated in terms of the amounts in issue, so I've used those instead) and, as far as I am aware, there is no freely available total return series for the UK stock market prior to that date. An argument could also be made for drawing the line at 1900 when the first dated gilt was issued (only undated gilts, or consols, were available before that) or 1915 when dated gilts were issued in reasonable quantities. So, practical issues may be more important.

cheers
StillGoing

I can't help much with UK stock returns before 1870 (which IIRC you mentioned in a previous post was yet another reason you weren't able to go back beyond 1870....lack of stock return data before 1870) but for bills/short-term "safe" instruments other than actual UK treasury bills how far back does you data set go? I have data on UK three month bank bill rates back to the early 1820s from NBER; failing that, if you want a different measure of a "risk-free" short term rate you could also use the BOE discount rate which FRED has monthly data to back to the mid-1840s and R. G. Hawtrey (in the book "A Century of Bank Rate" which is available free on Archive.org) has quarterly data for this rate back to the mid-1820s; this--combined with East India Company three and six month bill yields for the 1808 or 1809 to late 1810s/early 1820s period--was how Siegel and also Prescott and Mehra were able to assemble a "risk-free rate" series for the UK back that far.

[watchnerd]

Thank you for posting your article.
There are a couple of minor points, not extremely consequential per se, but they hint to what the philosophy of the article is as a whole.
- On Table 1 you report that the data set covers 226 single years, 222 5-year periods, 217 10-year periods, and so on.
That is true in a way, but for multiple year periods those periods are not independent from each other.
In fact, one could very well argue that the 50-year period from 1950 to 1999 and the period 1951-2000 are pretty much one and the same and you are double counting them. Over 200 years there are only four 50-year periods truly different from each other and of course there are a few more quite different from each other and many more almost identical to each other.
- My second point is when in the same table you write "odds that stocks beat bonds". Those are not odds, that is past data.
if we just stop at the 50-year period analysis, what you are showing is that earlier on bonds were better, then stocks and bonds were about the same, and finally stocks have been better (and this because of the non-independence of consecutive periods I mentioned earlier).
That's what happened; we do not know the odds of that happening.

Overall, you make a rather exhaustive analysis of historical data, cutting and dicing it every which way and that is valuable. What is missing is an hypothesis of a mechanism behind the fact that stocks are better (or that bonds are, or that the two are about equivalent).
But that mechanism is known and it points to stocks having a higher expected return. That is because bonds return is contractually set and it is quite decoupled from the state of the underlying business insofar debt is not defaulted.
That is the reason why investment grade bonds are little correlated to stocks, while junk bonds have a much higher correlation.
For the former, probability of default is very small and it does not matter if the issuer does very well, so-so, or badly: the bond holder will collect the agreed upon interest in almost all cases. Nothing more, nothing less.
The situation is different for junk bonds where there is a higher chance of default and issuers not doing very well will much increase their default rate, while issuers doing well will much decrease it, and default there is also a big factor in realized returns, in addition to the nominal yield.

On the other hand, the return from stocks does depend primarily on business results (although in a relative fashion, the stock of a company "doing well" in absolute terms may not do nearly as well if other companies do much better at the same time).
And that's all we need to know: if the two classes had the same expected return, but the return from bonds is much less uncertain, who would invest in stocks ? At most you would have few investors investing in stocks in the short term, when they need the chance of an extra return, but that chance would go to zero in the long term, so nobody in their sane mind would invest in stocks for the long term.
But investors do and, in most cases, they are not insane.

All of the above is rational if the cost of money is a constant.

But if liquidity is tight, the equilibrium changes, as the debt holders are able to make a larger claim on business proceeds.

And the same for labor -- if labor supply is tight, than the share of the business proceeds that accrue to labor vs capital can change.

(see Thomas Piketty's Capital)

[StillGoing]

Omnibus reply

I do want to authorize and legitimate all of you who find history irrelevant in an investment context. DO NOT read my paper—it has nothing to offer you.

In more measured terms, I can quite understand anyone who would reject returns from 1792 to 1842 as irrelevant—too ancient. The question that comes to mind is, Where do you draw the relevance line?

Some possibilities:
1. Yesterday’s price action?
2. Last month’s price action?
3. Trailing ten years only?
4. The advent of the smartphone economy in 2007?
5. 1926?
6. 1871?

You are free to answer: “I don’t look at the past at all, I proceed on fundamental theory.”

To which I would answer, “Is your theory falsifiable? And by what sort of data?”

And if your theory is not falsifiable by any data whatsoever … tell me how you know it’s a good theory?

McQ congratulations on the paper.

My position is that historical data offers useful guidance particularly, as others have said, for comparing different accumulation or decumulation strategies.

The question I might ask is "could the observed historical returns, particularly periods of poor performance, reoccur in a similar fashion in the future?"

For example,
1) Could high inflation affect real returns over an extended period of time (i.e., a rerun of the 1970s)?
2) Could the stock market crash by more than 50% over the course of a year or two (i.e., a rerun of 1929)?
3) Could real bond returns be low or negative over periods of a decade or more (a rerun of the 1940s onwards)?

My answer to each of these questions is 'yes', and this feeds in to my using historical returns as an indicative, but not predictive, tool. But it also argues for a more sophisticated analysis of such results than looking at a single number (e.g., the oft quoted 4% safe withdrawal rate) since 'optimising' for the retirement that provides the 'safe withdrawal rate' may have significant knock on effects on other retirements.

I'm currently working with a data set that goes back to the 17th century (UK government bonds) but have drawn a line for calculating returns at 1870 since this is the earliest date that UK treasury bills were issued (although from 1870 to about 1918, bankers bills predominated in terms of the amounts in issue, so I've used those instead) and, as far as I am aware, there is no freely available total return series for the UK stock market prior to that date. An argument could also be made for drawing the line at 1900 when the first dated gilt was issued (only undated gilts, or consols, were available before that) or 1915 when dated gilts were issued in reasonable quantities. So, practical issues may be more important.

cheers
StillGoing

I can't help much with UK stock returns before 1870 (which IIRC you mentioned in a previous post was yet another reason you weren't able to go back beyond 1870....lack of stock return data before 1870) but for bills/short-term "safe" instruments other than actual UK treasury bills how far back does you data set go? I have data on UK three month bank bill rates back to the early 1820s from NBER; failing that, if you want a different measure of a "risk-free" short term rate you could also use the BOE discount rate which FRED has monthly data to back to the mid-1840s and R. G. Hawtrey (in the book "A Century of Bank Rate" which is available free on Archive.org) has quarterly data for this rate back to the mid-1820s; this--combined with East India Company three and six month bill yields for the 1808 or 1809 to late 1810s/early 1820s period--was how Siegel and also Prescott and Mehra were able to assemble a "risk-free rate" series for the UK back that far.

Thanks for that - I've used prime bank bills to 1870 (since they were issued in far larger quantities than treasuries until just after WWI). Earlier data are also collected together in 'A millennium of macroeconomic history' (see https://www.bankofengland.co.uk/statist ... h-datasets) although, prior to 1870, they become monthly average (rather than end of month) and drop slightly down to the quality scale to 'prime and first class' rather than prime.

When I come back looking at gilts again (which will be shortly), I'll extend the data back to match the start of the stock series in the paper McQ linked above. What I was keen to establish in all of this was an open source set of returns for the UK with a bit more detail than a generic 'bonds' category (there are paid for versions of this).

cheers
StillGoing