I had been planning to start a thread on "Fixing the REHP Study" at some later date. However, the issue surfaced today in a post put forward by JWR1945 on the "Dead Posts File" thread, so I thought I should go ahead and open this thread now and move the discussion initiated by JWR1945 over here. I'll put forward some sketchy thoughts of my own now and add more detail to them at a later date.
To gain a full understanding of the context in which the JWR1945 post was put forward, you need to look at the discussion that takes place in the "SWR as a Tool" thread. JWR makes the point there that the 4 percent number reported as safe in the REHP study in fact is not safe according to the historical data. SalaryGuru put up a word game post in an apparent effort to trick community members who have not been following the debate closely into thinking that this is not so. I deleted the word game post from that thread, and placed the text of the deleted post (as well as an explanation of why it was deleted) in a new thread called "Dead Posts File."
JWR1945 offered the following comments in response to my thread-starting post:
I don't want us to discuss issues of substance in the "Dead Posts File" thread. That thread is reserved for the posting of the words of deleted posts and discussions of why the deletions were made. When I finish this post, I will put up a post on the "Dead Posts File" thread asking that discussions of the points raised by JWR1945 be moved here.I composed this earlier this morning and then found the original post missing:If I am not mistaken, Hocus is beginning to use the phrase Historical Database Survivability Withdrawal Rate (HDSWR) as a descriptive replacement for the older phrase Historical Database Rate (HDBR).The Historical Database Rates of the past were generally bounded by 3.9% or 4.0% (over 30 years and with 50% to 80% stocks).
We draw a sharp distinction on this board between Historical Database Rates, which tell us what would have survived in the past, and Safe Withdrawal Rates, which are mathematically calculated estimates of what will be safe in the future.
I think that it was a good decision to separate this from the original thread.
Have fun.
John R.
JWR1945 says in his post that "Hocus is beginning to use the phrase Historical Database Survivability Withdrawal Rate (HDSWR)...." Actually, the phrase that I use to describe the number reported in the REHP study is "historical surviving withdrawal rate (HSWR)." I am of course open to the use of any phrase that accurately describes the number reported in that study and other conventional methodology studies.
There are two points that must be conveyed by a phrase for it to be accurate:
1) It must be made clear to readers of the study that the methodology used in the REHP study was not designed to give us an indication of what withdrawal rate is likely to survive in the future. Since the study ignores a factor that has always in the past played a key role in the determination of what withdrawal rate is safe, there is no reason to believe that the number reported in the study is safe for retirements beginning today. I use a phrase including the word "historical" to alert readers to this critical limitation of the study.
2) It must be made clear to readers of the study that the methodology used in the REHP study was not designed to reveal the withdrawal rate that was safe at any time-period. The methodology employed identifies a withdrawal rate that in a hypothetical sense would have survived in earlier time-periods. It does not tell us whether the surviving withdrawal rate reported was a safe withdrawal rate at the beginning of any of the particular historical sequences examined. The reality is that, in some of the sequences examined, the 4 percent number survived the hypothetical test because it was safe (or even more than safe) while in others, it survived because it was lucky. It is not reasonable to presume that because a number survived once or twice for a historical sequence beginning from a specified valuation level that it is a safe withdrawal rate to use for other 30-year sequences of returns beginning from similar valuation levels. The study examines only the question of whether a withdrawal rate survived or not. It does not consider whether the rate survived because it was a safe number at the valuation levels that applied at the beginning of the retirement or whether it survived because it was a lucky number in one or two particular historical sequences examined.
Another approach to fixing the REHP study would be to add language pointing readers to the work that has been put forward discrediting it. In the event that intercst continues to this day to believe that his study provides an accurate report of what the historical data reveals re what withdrawal rate is safe , he is under an obligation to his readers to make them aware of the case that has been raised discrediting the study and of his reasons for not being persuaded by the case that has been made.
If this is the approach used to fix the study, he needs to include references to the claims made by William Bernstein that the conventional studies are misleading at times of high valuation and to Bernstein's determination that the REHP study was off the mark by a full two percentage points in the number it reported as safe at the top of the bubble. He also needs to make reference to the work that has been done at the various discussion boards, especially the work done at this board. Finally, he needs to discuss the claims that have been put forward by such experts as Rob Arnott, Robert Shiller, Andrew Smithers and others taking exception to the root assumption of the REHP study, that changes in valuation have zero effect on long-term returns and thus that no adjustment for changes in valuation is needed in a reasoned analysis of what withdrawal rate is safe.