Since I first published Part 7 of the SWR Series with the accompanying Google Sheet in early 2017, I’ve made several changes and enhancements. Sometimes without much explanation or documentation. So, it would be nice to do a quick update and itemize the changes since then. Whether this is the first time using the toolbox or you check it out again after more than a year, I hope you all find the new features useful…
Here again is the Google Sheet Link:
As always, please save your own copy because the current (clean) version posted on Google Sheets has to be write-protected so visitors don’t mess around with my formulas! You will be asked to create your own personal copy, which you can edit! 🙂
Main Tab: more detailed results
As before, you enter almost all parameters in the main tab “Parameters & Main Results.” All fields in Orange are user inputs, such as:
- Equity/Bond/Cash/Gold share. To make sure the weights sum up to 100%, the Gold share is set to 100% minus the sum of the rest.
- Fama-French style factors. This is new, see item #3 below for more details.
- Assumed projected asset returns for the asset classes going forward. Why do this? I’d like to be able to simulate 60-year windows of the 1960s and 1970s retirement cohorts and don’t want to be constrained by a slightly shorter than 60-year data availability. Since the SWR is overwhelmingly determined by the first 10-15 years of the retirement horizon (Sequence Risk, see SWR series Part 15), the last few years of “made up” returns during retirement don’t have much of an impact on the SWR estimates. Example: the failsafe WR is 3.46% during the 1960s with our assumed 3.75% p.a. equity return. If we increase the equity expected return to 20% p.a., the failsafe increases to only 3.52%. But, of course, the 2000 and especially 2007 fail-safe numbers would be significantly impacted by the return forecasts. So, use those figures with a grain of salt!
- The retirement horizon in months and the target final value (e.g. for bequests) as a percentage of the initial value.
- Notice that the supplemental cash flows are no longer inputted here in this main tab, but in a separate tab, see item #2 below.
The main results table is pretty large so I split it in two, the first part is below:
- As before, I display the failure rates of different initial withdrawal rates in the top half of the table.
- The bottom portion of the table is new and presents the calculation the other way around: Pick a desired failure rate and look up the withdrawal rates to match it. The first row in that portion at 0.00%, i.e., this is the fail-safe initial withdrawal rate. But you can also look for other failure rates, 1%, 2%, 5%, and 10% as well and see how much extra initial withdrawal you could have sustained in historical simulations. And I hope that nobody would even consider a strategy with a failure rate of 25 or even 50%, but I display those just for reference.
- The columns calculate the corresponding stats for all retirement starting months and years since 1926 (the start of the Fama-French database and also the starting point for the Trinity Study) and 1950 and also for different Shiller CAPE Ratio regimes (under 20 vs. 20 to 30 vs. 30+). Notice that as of late August 2018, we’re at over 32! As conservative as I am normally, though, I’d concede that today’s elevated CAPE ratio doesn’t quite “feel” like the 30+ CAPEs of the past. Maybe I’d still consider the current CAPE a high 20s, not really a low 30s!
- Notice how sensitive the failure rates can be when changing the WR in 0.25% steps, especially when the CAPE is high!
And Part 2, see below. The table has the same format as the first part, but now the columns are conditional on how far the S&P 500 index is away from its most recent high. Why does this matter? The stock market is a Random Walk and past returns have no bearing on future returns, right? Wrong! As is well-known in finance and as I pointed out in a post a few months ago, stocks have the tendency to mean-revert. So, expected returns tend to be higher after a steep drop and lower after a long run-up in stock prices. That’s reflected in the failure probabilities of the 4% Rule below. The unconditional failure probability was just under 10%, but it was 18.55% when we were at the market peak. So, failure probabilities are indeed impacted by past returns. File this as another piece of evidence that the stock market isn’t exactly a Random Walk!
Why is this relevant? Well, on August 22, 2018, the market just became the longest-running bull market in history (though not everyone agrees on the definition of the longest bull market, see this article on Bloomberg). If history is any guide the retiree in this numerical example will be wise to be more cautious with the initial withdrawal rate and not just blindly apply the 4% Rule mantra. Maybe withdraw a little bit less, 3.25-3.50% to have a bit a cushion. In other words, after a potential drop in the portfolio the effective withdrawal rate might reach 4% right around the time when the S&P had a ~20% drop, which is when, historically, the 4% becomes very safe again!
I also added a small table with the fail-safe withdrawal rates in five important time intervals around stock market peaks. That’s because I usually like to see during what period the all-time fail-safe occurred. Most of the time it’s either 1929 or the 1960s, depending on the stock vs. bond. vs. cash allocation! It’s always amazing to see that the 1970s and early 80s recessions did an even worse trick on the mid-60s retirement cohorts than the ones that retired around the 1973 market peak!
Supplemental cash flows: now easier to input!
One of the biggest challenges for folks trying to use this spreadsheet used to be how to correctly enter the supplemental cash flows. I tried to make this a little bit easier and more intuitive, so I created a separate tab, appropriately named “Cash Flow Assist” to help with that. At the top, we can input the initial portfolio value and a projected inflation rate to discount the value of any non-inflation-adjusted future cash flows, e.g., corporate pensions, etc. in the orange fields below. As always, it’s best to go through a simple example:
- The portfolio is worth $3m at the beginning of retirement.
- We expect 2% inflation going forward.
- Spouse 1 expects Social Security after 25 years (=month 301) worth $2,000 per month.
- Spouse 2 expects Social Security after 26 years (=month 313) worth $800 per month.
- Spouse 1 expects a small corporate pension of $300 (not inflation-adjusted) 11 years into retirement.
- We expect $1,000 in additional monthly expenses (e.g. medical) 30 years into retirement (in today’s dollars, adjusted for inflation) and that amount rises to $2,000 after 40 years. This corresponds to the two spouses reaching a certain age where they may scale back other expenses (travel) but face an increase in medical, in-home care expenses, etc. which will case in a net increase in expenses.
How do we input all of this? At the top of the page we input the net worth and an inflation figure and leave the other orange fields empty (=$0) for now because there are no supplemental cash flows during the first year, see below:
Next, we input the corporate pension, starting after 11 years (month 133), see below. Notice that we input this in the fifth column where the non-COLA cash flows reside. So, this will be discounted to make sure it’s comparable to today’s real dollars!
Next, the Social Security payments start 25 and 26 years into retirement, see below. The benefits are inflation-adjusted so they are entered in the corresponding columns with COLA cash flows:
And, finally, the additional budget for medical expenses, care, nursing homes, etc., see below. This is now inputted as a negative cash flow!
In the “green” column on the right, the program translates the different cash flows into percentages of the initial portfolio:
Also in the same tab is the same table that showed up in the main tab but the percentages are translated into withdrawal amounts in dollars p.a.:
Fama-French Style Factors (since 1926)
I added that feature pretty early in the spring of 2017, mostly out of curiosity about how much of a difference some of the widely cited style premia such as value and size would have made. See Ken French’s site for the data and more documentation on the construction of the Fama-French factors. For example, let’s simulate the following scenarios:
- The 80% S&P500 baseline
- Keep the overall equity share at 80%, but use a 25% small caps and 25% value stocks tilt: Set the SMB and HML allocation to 25% each. Make sure you keep the overall stock allocation at 80%!
- Keep the overall equity share at 80%, but use 50% growth stocks. Set the HML allocation to -50%. Again, make sure you keep the overall stock allocation at 80%!
The small-cap plus value bias would have easily lifted the SWR to above 4%, see table below. But I’d probably not get too excited about this result. There’s no guarantee that the size and value premium will persist forever and continue to help future retirees. All this could just be backward-looking bias. Certainly, in 1926 nobody would have known about the work by Fama and French. In fact, if you had been wrong and bet on the wrong style, for example, “growth” instead of “value” you would have totally ruined your safe withdrawal rates. Fail-safe withdrawal rates are now in the low 2% range. Ouch!
Case Study: glide path simulation
In the tab “Case Study” where you can simulate one single time series of the portfolio values for a specific starting date and initial withdrawal rate, I also added a simple glidepath simulation for comparison (see Part 19 and Part 20 of the SWR Series). This is the simplest possible version with just the static glidepath going between two different equity weights in fixed steps (e.g. 0.3% per month) and investing the residual in bonds, see below:
Compared to an 80/20 static stock/bond allocation, the glidepath would have made a huge difference during the Great Depression! But glidepaths were not a panacea because, during the 1970s, bonds offered much less diversification.
Simulate CAPE-based Withdrawal Rules
I added another tab to simulate CAPE Rules with different parameters. Just as a recap, the CAPE-based rule, in its simplest form, expresses the annualized target withdrawal rate as a+b times the inverse of the Shiller CAPE (=CAEY = Shiller Earnings yield). See Part 18 for more details and why I like this approach. My preferred rule would be to set the intercept to around 1.5-1.75% and the slope to one-half. In the example below I use 1.75%/0.50. Notice that a constant percentage rule would be a special case if we set a=4% (or whatever rate you like) and b=0, i.e., withdraw 4% p.a., regardless of equity valuations.
Running out of money is no longer an issue with the CAPE-based rules. Failure comes in the form of deep and extended cuts to consumption. So, to compare how much (or how little) I like different CAPE I tend to look for 3 key stats:
- What’s the change in real, inflation-adjusted withdrawal amounts over a 30-year horizon?
- Notice that the point-to-point comparison over 30 years can deep cuts in spending, so I also like to know the lowest withdrawal amount relative to the initial withdrawal amount
- And finally, what’s the average over the 30-year window relative to the initial amount?
See below for a numerical example:
I’m interested in how the three measures would have evolved in the worst-case scenarios, so I calculate them for different time intervals (all months vs. 1926 onward) as well as the worst-case scenarios in the five different “troublemaker” retirement cohorts (Great Depression, the 1960s, early 1973, Dot-Com bust and Great Recession), see the table below. I also add the volatility of year-over-year withdrawals and some stats on the withdrawal rates implied by this CAPE rule:
A little side note: I frequently get questions and comments about the CAPE parameters just like recently when a reader wondered why I assign a weight of 0.5 on the CAEY. Shouldn’t the withdrawal rate be less volatile with a higher intercept and lower slope? Yes, but as a retiree, I’m less worried about volatility in withdrawal rates and more worried about volatility in withdrawal amounts. So if I were to set the intercept to 3% and the slope to 0.3 I’d have a more volatile stream of withdrawals, see below. And the more you reduce the slope parameter the closer you get to the good old constant percentage rule where your withdrawals become just as volatile as the portfolio itself!
I guess it’s up to you and what you feel comfortable with but I like the way the CAPE rule cushions the volatility in withdrawal amounts, see below:
OK, so much for today! I hope you enjoy the new features! Please let me know if you find bugs or like to suggest more features!
Thanks for stopping by today! Please leave your comments and suggestions below! Also, make sure you check out the other parts of the series, see here for a guide to the different parts so far!
Also notice, all the usual disclaimers apply!
Picture Credit: Pixabay