In an attempt to understand unexpected CAPE Rules results, I'm comparing CAPE Rules (CAPE a=3%, CAPE b=0) with Fixed % Spending (Spending Rate=3%), both using Historical Results.
First the unexpected CAPE results, again CAPE a=3% and CAPE b=0, historical:
The Success Rate in the final year is 65%, and the Worst Historical Start Year is 1973. However, when I activate the Historical Sequence Analysis for 1973, the Savings line does not got to zero--on the contrary, it shows a significant savings amount remaining in the final year. I'm struggling to understand why the overall Success Rate is so poor when the worst year still succeeds by a huge margin.
To try to understand this I ran a Fixed % Spending analysis with Spending Rate=3%, on the assumption that this is equivalent to CAPE a=3% and CAPE b=0:
In this analysis the Success Rate is 100% for all of the final ten years, and the Worst Historical Start Year is 1937, which still succeeds by a healthy margin but not as much as the 1973 result in the CAPE analysis above.
I got to this point when I first saw unexpected results with different CAPE variables (e.g., a=1%, b=0.5), so I eliminated the CAEY variability by setting b=0. When that still showed unexpected results, I decided to compare to the Fixed % results, which AFAIK should be the same as CAPE a=x%, b=0.
Needless to say, I'm struggling to understand these results and would appreciate any help or insight.
Just a note here on the seminal references on using the CAPE-based rules from the originator (Karsten Jenke): https://earlyretirementnow.com/
Jenke offers interesting and useful spreadsheet tools (for free), and among them is one on the calculation of the safe withdrawal rate ("SWR"):
https://earlyretirementnow.com/2018/08/29/google-sheet-updates-swr-series-part-28/
(1) This link is mentioned in the Pralana manual. A sensitivity analysis to the "a" and "b" parameters is provided in this post (but does not actually use Pralana per se) and recommends a=1.75% and b=0.5 for the portfolio studied to establish an initial CAPE-based SWR. The object here is to tie smoothed equity (S&P 500) earnings (10-year CAPE) to the portfolio withdrawal rate for a "smoother ride" and avoid running out of money.
(2) This link considers modifying the Shiller CAPE for modern corporate fiscal behavior (stock buy-backs, lower corporate tax rates, lags in Shiller data...). The alternative CAPE offered seems to generally be lower than the Shiller value currently (which implies a slightly higher SWR - currently about 3.6% vs about 3% using the most recent Shiller CAPE value).
https://earlyretirementnow.com/2022/10/05/building-a-better-cape-ratio/
Frequent updates of the alternative CAPE is offered (almost daily) at: https://drive.google.com/file/d/1ugtRN3TaAVwQi-20mjt4DctF-glppSMD/view?usp=sharing
(3) This link shows an analysis of the SWR using the proposed modifications to the Shiller CAPE value using the SWR tool mentioned above (with a=1.75% and b=0.5)
(4) This link provides a critique of assuming alternative flexible SWRs using the tool mentioned above:
https://earlyretirementnow.com/2023/06/16/flexibility-swr-series-part-58/
I am not clear if the detailed implementation of the CAPE-based approach in Pralana (which uses the Shiller CAPE data) is completely consistent with Jenke's analysis approach. One notes, for example, that Pralana is using annual CAPE values, while the SWR tool uses monthly data (argued by Jenke as more relevant to retiree modeling of the SWR).
Anyway, I find that (with a=1.75% and b=0.5) the WR produced by Pralana with CAPE-based rules (Historical Analysis mode only) appears to be generally similar to that from the Jenke SWR-tool approach.
@wallace471 Thanks for sharing the details surrounding Karsten Jeske's (i.e, "Big ERN's") CAPE method. I've read all the articles above and use his planning spreadsheet to augment the insights I get from Pralana, but I'm sure others will find the background and links useful if they haven't already explored this material.
In fact, my question above is based on Karsten's observation: "Notice that the constant percentage rule [Pralana calls this Fixed %] is simply a special case of the CAPE-rule if we set b=0 and a=4% (or whatever your desired constant percentage may be)." This is why I'm confused by the differences between Pralana's results for a Fixed % of 3% and CAPE with a=3%, b=0. Note that the actual percentage is irrelevant as long as the same value is used in both spending strategies.
One big difference between Karsten's model and Pralana's is that Pralana is only applying the CAPE spending strategy (or any other of the spending strategies) to your variable (non-essential) spending, where Karsten's model applies to all spending, although there is some rudimentary capability to include future income and expenses in Karsten's spreadsheet. This means that it's hard to directly compare the models using the same spending strategy parameters. For example, using CAPE a=1.75, b=0.5 in Pralana only applies to variable spending on top of your explicitly specified essential spending, while Karsten's model applies CAPE a=1.75, b=0.5 to the entire spending envelope. The same is true for fixed/constant and other spending strategies.
I like Pralana's approach, but it does make it hard to compare to more simplistic models like Karsten's CAPE model, the "4% rule", or other well known models, all of which simply look at overall spending. Rather than trying to compare directly, I use Karsten's model to help validate the big picture, and I use Pralana to do the detailed planning.
Are there any other thoughts on the original question?
