KenM: I could never buy into any study or any analysis with complete certainty for the next 30 years
JWR1945: hocus is burdened by the ghosts of the recent past. This specific proposition has been offered repeatedly (elsewhere) as being absolutely certain...absent the most extreme of unlikely circumstances.
You are a little off with this comment, in my view, JWR1945. I will try to work down the logic chain a bit.
When I say that Bernstein puts the odds of a future worse than the past at 5 percent, I am going by what some people quoted him as saying on the other board. I have not looked at the quote myself and it is possible that the interpretations of it that I read were inappropriate. I am not at all dogmatic in the view that the chance of the future being worse than the past is 5 percent. It could be 10 percent or 15 percent or 20 percent or whatever. But I don't see that it matters all that much for practical purposes.
Let's say that the chances of the future being worse than the past is 20 percent rather than 5 percent. I still don't see that there is anything that I can do about it. What sort of event would be worse than any that has occured in the past 130 years? Nuclear war? What am I going to do about it if I think that there is going to be a nuclear war, increase my allocation to TIPS? How is that going to help me?
The point we are discussing here is one in which I am more or less in agreement with intercst. He states things in a more dogmatic fashion than I would, but I think that his basic approach on this particular point is generally sound. He acknowledges that the future may be worse than the past, but he argues for making your investment decisions based on the premise that it will not be. I am to a large extent sympathetic to that argument.
I think that what is confusing people is that people are thinking that stock returns that are lower than we have seen in the past constitute a case of the future being worse than the past. This is a mistaken interpretation of the caveat, in my view. If the caveat referred to lower stock returns, SWR claims would be a tautology. The claim would be: "The future will be like the past unless the future is not like the past." That is such a silly claim that it rules out interpreting the caveat in that way, in my view.
The more reasonable way to interpret the caveat is the way in which intercst always interpreted it prior to May 13, 2002. The caveat refers to terrible events outside of the sorts of things considered in the SWR analysis. The caveat does not refer to the possibility that returns will be worse than the past for normal sorts of reasons. That possibility is what the analysis is aiming to examine. It refers to the possibility that returns will be worse for entirely different sorts of reasons, catastrophic developments that never took place in the 130 years, like nuclear war or political instability.
The reason why the possibility of these sorts of events is covered in a caveat is that there is no way to assess the possibility of them occuring by making reference to data. They must be covered by a caveat because the SWR analytical process has no means by which to address them. They are outside the reach of the analytical process being set up.
You have two sorts of things with a potential to affect your returns. Investment-type things that can be measured and that should be incorporated into the SWR analysis. And non-investment type things that cannot be measured and which are not incorporated into the SWR analysis. As a general rule, there is nothing that can be done about things in the latter category anyway, so, even if you think that the odds of them turning up are greater than 5 percent, it does not necessarily influence your allocation choice to think that.
Things get sticky when you consider investment type things that cannot be incorporated into the SWR analysis.
Consider the demographics issue that has been raised from time to time as an example of something other than nuclear war that could make the future worse than the past. If you can show that demographics changes have affected stock returns in the past, and if you can gather reasonably good data on how they are likely to affect stock returns in the future, such data should be included in your SWR analysis. That's the best way to address this concern.
Let's say that you cannot gather reasonable data on this point, so you leave it out of your analysis. That's unfortunate, but it does not necessarily change your decisions re your take-out number or re your allocation percentage. If you do not have enough data to include the factor in the analysis,. you probably will not be able to say whether this factor is going to cause you to want to go with a lower stock allocation or not. If the factor does not affect your investment decisions, it does not have any practical significance re the purpose of SWR analysis--crafting a FIRE plan.
Pre-May 13, 2002, intercst defined the caveat to cover "things that cannot be calculated and which thus cannot be incorporated into the study" (those are my words to describe his position, not his own). I think that is a generally sensible way to go with the caveat. If you can measure it, include it. If you cannot, leave it out. But if you leave it out because you cannot measure it, don't worry too much about it because there probably is not a good way to use the factor in making investment decisions anyway.
Where it gets subjective is when you can measure the effect of the factor to some extent, but not to a satisfactory extent. What if you were sure that the demographics issue would cause some drop in long-term returns but not able to make much more than an educated guess as to how much? In that event, I think that the thing to do is to say that studies incorporating the factor are valid and studies not incorporating the factor are valid, but both types of studies should take note of how the factor is treated in the write-up section of the study. There is just not enough certainty on the point to permit dogmatism.
That does not mean, however, that there is not enough certainty on other points to justify some dogmatism on them. On the question of valuation levels, I am being dogmatic on the point that an adjustment of some sort must be included. I don't think it is clear which sort of adjustment is best, so I am not being dogmatic on the how-to-do-it question.
The goal should be to produce the most objective measure of the SWR possible. In essence, the analytical technique is not a rule of thumb. It aims at scientific accuracy. It fails to achieve this goal only because we do not possess access to perfect knowledge of what affects the result and we do not possess perfect data on the factors that we know do so. But methodology questions should still be addressed in such a way as to reflect the purpose of the tool, which is to provide an objective assessment of risk.
Bernstein is right to be skeptical of claims of 100 percent certainty or 95 percent certainty made in regard to existing studies. But that is largely because existing studies leave out several factors that bear on the result. If you included data on as many factors as possible, you would have a much more valid tool. Perhaps it would not provide 95 percent certainty. I don't know what the correct odds are that the future will be worse than the past. But if a study includes all the factors known to affect the result and for which data is available, doesn't it include pretty much all the factors that you could take into consideration in determining your take-out number and your allocation percentage anyway?
What I am looking for is an analytical tool that takes all the data that should go into deciding these two issues and makes use of them in an analytically valid way to transform a decision-making process that has long been a subjective one into a largely objective one. The thing that I like about SWR analysis is that it is largely not guesswork. Some guesswork is unavoidable because of the imperfection of the data available to us. But the whole idea is to limit the guessing part of the equation to the greatest extent possible.
We cannot make the tool a perfectly objective assessment of risk factors because we do not possess access to all the data needed to do so. But why not make it as objective as it can be, based on what we know today about the realities of SWRs and on what data is available to us to do so?