Or were you just saying you think the US was exceptional during the 20th century, and investors should not expect similar returns
Yes, that is what I am saying.
whether that particular chart took into account survivorship bias
I meant survivorship bias in the country sense. What’s the return of a German stock portfolio over the last century? It is zero—the portfolio went to zero in WW2 and without additional money invested it stays at zero.
On the one hand, this is an important issue and shouldn’t be ignored if you’re planning for your retirement.
On the other hand… Let’s think about a scenario where you’ve worked hard and saved hard until (say) the age of 40, and then 10 years later there’s a national catastrophe on the level of losing a major war which wipes out all your savings. You are, indeed, going to be in trouble. But so is someone who’s been working for pay all that time: they’ve lost all their savings too, and probably their job. Either position’s going to be pretty terrible.
But if your retirement portfolio is internationally diversified (and it should be!) then you aren’t just vulnerable to war and revolution in your home country, you are vulnerable to war and revolution in any of the countries where you are invested. Survivorship bias is definitely relevant.
Sure. But now I remark that there are lots of countries and such total wipeouts are really quite rare. So, e.g., if your portfolio is something like equally divided among 10 major countries, and each of them has a total wipeout once per 30 years (of course these are both really crude approximations), then what happens is that once per 30 years you lose 10% of your investments, which is kinda like losing 0.3% per year, which is about what most index funds charge in management fees. (Of course it’s worse really because it’s “lumpier”.)
So, again, it’s an important issue but I remain to be convinced that it’s that important an issue.
if your portfolio is something like equally divided among 10 major countries, and each of them has a total wipeout once per 30 years
Why don’t you look at reality instead of going for abstract approximations? You are assuming that a “major country” being wiped out by a war would not affect other countries. Really? The 2008 crisis didn’t even come close to being a wipeout and how correlated were the stock markets of the major countries during the crisis? Or, if you want to go back to WW2, which stock markets remained unscathed while Germany was wiped out?
I am concerned that our argument takes the form: “I think this effect is bigger than you think it is!” “Oh, but I think it’s smaller than you think it is!” repeated a few times. For all we know, we agree about the actual effect but have wrong ideas about one another’s estimates of how big it is. Can we perhaps find some actual concrete proposition that we disagree on? (Or, as may well happen, find that there isn’t one.)
Here’s a candidate. It’s a more detailed and, where possible, quantitative statement of my own opinion. It’s a conjunction so there should be plenty to disagree with :-). I don’t believe anything I’ve said in this discussion has been incompatible with any of these points.
(1) Survivor bias at the country level is an important issue and you shouldn’t ignore it in estimating the prospects of an investment portfolio. (2) The possibility that your own country gets wiped out economically is real; it has probability on the order of 0.5% or so per year (let’s say 0.25% to 1.0%), it affects people who are still working at least half as badly as early retirees (I suspect nearer to parity than that), and while it’s worth trying to be prepared for it I don’t see it as a major factor in deciding whether to retire early. (3) The possibility that another country you’re heavily invested in gets wiped out is also real; if your international investments are reasonably diversified (or small, though that has its own problems) then the impact is on the order of 0.3% loss per year. More precisely, I’ll say 0.2% to 1.0%. (4) The possibility that another country with impact on one you’re heavily invested in gets wiped out is also real; there are more ways for it to happen but the impact is smaller per instance. However, it’s somewhat “priced in” already even for someone who’s ignored the issue, because such impacts already affect whatever indices they’ve looked at. Let’s say that the size of this effect beyond what a naive prospective early retiree is already expecting is in the range 0% to 0.5% depending, e.g., on where they are. (5) The net effect of all this is that if you’re considering retiring early, you should deduct 1% or so from your estimate of annual investment growth if you haven’t previously contemplated this sort of risk, and when comparing the risks of early retirement with those of continuing to work you should not forget the risk of your own country getting wiped out economically by war, hyperinflation, etc.
Why don’t you look at reality instead of going for abstract approximations?
Because doing the necessary research would take maybe 10x as long (quite possibly 100x as long) and I have higher-priority uses for that much time.
You are assuming that a “major country” being wiped out by a war would not affect other countries. [...]
No. I am assuming that what we’re interested in here is the differential impact of this sort of disaster on an early-retired person’s savings, compared with its impact on a still-working person’s job, job prospects, and savings. (Because the relevant question is whether retiring early and living on savings is a worse decision than it might otherwise look like, because of this danger.)
The more countries affected by a disaster, the more likely it is that there’s a big effect on the country in which you live, and hence the more likely that it hits still-working people hard too.
And yes, countries’ outcomes are correlated. Also, there are considerably more than 10 countries of substantial size in the world. Using a low figure there serves as a partial correction for the correlation, as well as for the fact that most people don’t have a portfolio balanced equally across all the world’s countries.
The historical performance of an index like the S&P or Nikkei has the cross-country effects of disasters elsewhere already factored in. Such effects will be underestimated by someone looking at historical index performance only if the indices they look at are, for some reason, less affected by cross-country effects than others relevant to them. That certainly might be true (e.g., European countries’ stock markets may be more correlated with one another than any of them is with the US, in which case looking at the DJIA or the S&P 500 would underestimate the risk of being hit by another country’s disaster if you’re in the US but hold substantial investments in Europe) but this feels like a second-order effect to me.
I think my original point was just that the effect (survivorship bias at the country level) exists and most people happily ignore it. Looks like we agree about that.
My follow-up point was that this effect—survivorship bias at the country level—belongs to the class of things which makes your estimates of future returns suspect. However we’ve moved on to another issue—how to account for the possibility of a major disaster (war, revolution, hyperinflation, etc.) while planning your life and what does this possibility and its consequences look like.
I think I’d like to stress the the consequences will be complicated and widespread, not reducible to lopping a percentage point off your expected returns. I also think that estimates have to be country-specific. The probabilities of the US going down are noticeably less than the probabilities of, say, Russia, going down. However a Russian implosion will be more contained (in the sense of affecting other countries) while if the US implodes it might well take the entire North America and Western Europe down with it.
One additional point is that you’re not only interested in the expected return on your portfolio. You are also interested in the expected variance. The probability of disaster does not only reduce your expected return, but it also pushes up, considerably, your expected variance as well as makes your expected probability distribution asymmetric (more asymmetric, really).
By the way, I do not agree that “the historical performance of an index like the S&P or Nikkei has the cross-country effects of disasters elsewhere already factored in.” The reason is that the interdependency of countries is not a constant. It grew a LOT during the XX century, especially its last few decades. For example, S&P is much more correlated with Nikkei now than it was in the 60s. Chinese economic numbers whipsaw Brazil (which exports huge amounts of iron ore to China) and significantly affect the US stock market. Or, remember what happened to the US stock markets when it turned out that Greece is bankrupt?
The world is much more interrelated now than it used to be. Historical performance does not reflect the current reality.
Let’s think about a scenario where you’ve worked hard and saved hard until (say) the age of 40, and then 10 years later there’s a national catastrophe on the level of losing a major war
These are different issues.
This subthread is basically about estimating future returns from diversified stock portfolios and whether S&P returns for the last few decades provide a good baseline for that.
You are talking about the stability of life and about whether saving money is useful if there’s a chance your country will be smashed into little bits.
By the way, a much more likely scenario for a Western country is not losing a major war but having a hyperinflation episode. In this case the guy with the savings loses all, while the guy with a job is much better off.
You raised the issue of survivorship bias at the country level and gave the example of a country wiped out by a major war. So I explained why, if you’re adjusting the expectations of a retiree to account for what that sort of event could do to their investments, you also need to adjust the expectations of a non-retiree, who will also be hit hard by it.
Hyperinflation is indeed a good example of something that could hurt the retiree a lot worse than someone still working, but it seems to me that it depends a lot on (1) what form the retiree’s savings take and (2) what causes and consequences the hyperinflation has. For instance, if investments in the stock market lose a lot of their (real) value in a hyperinflationary episode, I’d expect that to be accompanied by a lot of job losses—so the worst case for a retiree with a lot of stock-market investments is also bad for someone still working.
Yes, that is what I am saying.
I meant survivorship bias in the country sense. What’s the return of a German stock portfolio over the last century? It is zero—the portfolio went to zero in WW2 and without additional money invested it stays at zero.
On the one hand, this is an important issue and shouldn’t be ignored if you’re planning for your retirement.
On the other hand… Let’s think about a scenario where you’ve worked hard and saved hard until (say) the age of 40, and then 10 years later there’s a national catastrophe on the level of losing a major war which wipes out all your savings. You are, indeed, going to be in trouble. But so is someone who’s been working for pay all that time: they’ve lost all their savings too, and probably their job. Either position’s going to be pretty terrible.
But if your retirement portfolio is internationally diversified (and it should be!) then you aren’t just vulnerable to war and revolution in your home country, you are vulnerable to war and revolution in any of the countries where you are invested. Survivorship bias is definitely relevant.
Sure. But now I remark that there are lots of countries and such total wipeouts are really quite rare. So, e.g., if your portfolio is something like equally divided among 10 major countries, and each of them has a total wipeout once per 30 years (of course these are both really crude approximations), then what happens is that once per 30 years you lose 10% of your investments, which is kinda like losing 0.3% per year, which is about what most index funds charge in management fees. (Of course it’s worse really because it’s “lumpier”.)
So, again, it’s an important issue but I remain to be convinced that it’s that important an issue.
Why don’t you look at reality instead of going for abstract approximations? You are assuming that a “major country” being wiped out by a war would not affect other countries. Really? The 2008 crisis didn’t even come close to being a wipeout and how correlated were the stock markets of the major countries during the crisis? Or, if you want to go back to WW2, which stock markets remained unscathed while Germany was wiped out?
I am concerned that our argument takes the form: “I think this effect is bigger than you think it is!” “Oh, but I think it’s smaller than you think it is!” repeated a few times. For all we know, we agree about the actual effect but have wrong ideas about one another’s estimates of how big it is. Can we perhaps find some actual concrete proposition that we disagree on? (Or, as may well happen, find that there isn’t one.)
Here’s a candidate. It’s a more detailed and, where possible, quantitative statement of my own opinion. It’s a conjunction so there should be plenty to disagree with :-). I don’t believe anything I’ve said in this discussion has been incompatible with any of these points.
(1) Survivor bias at the country level is an important issue and you shouldn’t ignore it in estimating the prospects of an investment portfolio. (2) The possibility that your own country gets wiped out economically is real; it has probability on the order of 0.5% or so per year (let’s say 0.25% to 1.0%), it affects people who are still working at least half as badly as early retirees (I suspect nearer to parity than that), and while it’s worth trying to be prepared for it I don’t see it as a major factor in deciding whether to retire early. (3) The possibility that another country you’re heavily invested in gets wiped out is also real; if your international investments are reasonably diversified (or small, though that has its own problems) then the impact is on the order of 0.3% loss per year. More precisely, I’ll say 0.2% to 1.0%. (4) The possibility that another country with impact on one you’re heavily invested in gets wiped out is also real; there are more ways for it to happen but the impact is smaller per instance. However, it’s somewhat “priced in” already even for someone who’s ignored the issue, because such impacts already affect whatever indices they’ve looked at. Let’s say that the size of this effect beyond what a naive prospective early retiree is already expecting is in the range 0% to 0.5% depending, e.g., on where they are. (5) The net effect of all this is that if you’re considering retiring early, you should deduct 1% or so from your estimate of annual investment growth if you haven’t previously contemplated this sort of risk, and when comparing the risks of early retirement with those of continuing to work you should not forget the risk of your own country getting wiped out economically by war, hyperinflation, etc.
Because doing the necessary research would take maybe 10x as long (quite possibly 100x as long) and I have higher-priority uses for that much time.
No. I am assuming that what we’re interested in here is the differential impact of this sort of disaster on an early-retired person’s savings, compared with its impact on a still-working person’s job, job prospects, and savings. (Because the relevant question is whether retiring early and living on savings is a worse decision than it might otherwise look like, because of this danger.)
The more countries affected by a disaster, the more likely it is that there’s a big effect on the country in which you live, and hence the more likely that it hits still-working people hard too.
And yes, countries’ outcomes are correlated. Also, there are considerably more than 10 countries of substantial size in the world. Using a low figure there serves as a partial correction for the correlation, as well as for the fact that most people don’t have a portfolio balanced equally across all the world’s countries.
The historical performance of an index like the S&P or Nikkei has the cross-country effects of disasters elsewhere already factored in. Such effects will be underestimated by someone looking at historical index performance only if the indices they look at are, for some reason, less affected by cross-country effects than others relevant to them. That certainly might be true (e.g., European countries’ stock markets may be more correlated with one another than any of them is with the US, in which case looking at the DJIA or the S&P 500 would underestimate the risk of being hit by another country’s disaster if you’re in the US but hold substantial investments in Europe) but this feels like a second-order effect to me.
I think my original point was just that the effect (survivorship bias at the country level) exists and most people happily ignore it. Looks like we agree about that.
My follow-up point was that this effect—survivorship bias at the country level—belongs to the class of things which makes your estimates of future returns suspect. However we’ve moved on to another issue—how to account for the possibility of a major disaster (war, revolution, hyperinflation, etc.) while planning your life and what does this possibility and its consequences look like.
I think I’d like to stress the the consequences will be complicated and widespread, not reducible to lopping a percentage point off your expected returns. I also think that estimates have to be country-specific. The probabilities of the US going down are noticeably less than the probabilities of, say, Russia, going down. However a Russian implosion will be more contained (in the sense of affecting other countries) while if the US implodes it might well take the entire North America and Western Europe down with it.
One additional point is that you’re not only interested in the expected return on your portfolio. You are also interested in the expected variance. The probability of disaster does not only reduce your expected return, but it also pushes up, considerably, your expected variance as well as makes your expected probability distribution asymmetric (more asymmetric, really).
By the way, I do not agree that “the historical performance of an index like the S&P or Nikkei has the cross-country effects of disasters elsewhere already factored in.” The reason is that the interdependency of countries is not a constant. It grew a LOT during the XX century, especially its last few decades. For example, S&P is much more correlated with Nikkei now than it was in the 60s. Chinese economic numbers whipsaw Brazil (which exports huge amounts of iron ore to China) and significantly affect the US stock market. Or, remember what happened to the US stock markets when it turned out that Greece is bankrupt?
The world is much more interrelated now than it used to be. Historical performance does not reflect the current reality.
These are different issues.
This subthread is basically about estimating future returns from diversified stock portfolios and whether S&P returns for the last few decades provide a good baseline for that.
You are talking about the stability of life and about whether saving money is useful if there’s a chance your country will be smashed into little bits.
By the way, a much more likely scenario for a Western country is not losing a major war but having a hyperinflation episode. In this case the guy with the savings loses all, while the guy with a job is much better off.
You raised the issue of survivorship bias at the country level and gave the example of a country wiped out by a major war. So I explained why, if you’re adjusting the expectations of a retiree to account for what that sort of event could do to their investments, you also need to adjust the expectations of a non-retiree, who will also be hit hard by it.
Hyperinflation is indeed a good example of something that could hurt the retiree a lot worse than someone still working, but it seems to me that it depends a lot on (1) what form the retiree’s savings take and (2) what causes and consequences the hyperinflation has. For instance, if investments in the stock market lose a lot of their (real) value in a hyperinflationary episode, I’d expect that to be accompanied by a lot of job losses—so the worst case for a retiree with a lot of stock-market investments is also bad for someone still working.
Hmm, interesting points. I had not seriously taken into account survivorship bias in this national sense before. I will have to think more about that.