Home » Blog » Deciding When It’s Safe to Retire Early (Part 2)

Deciding When It’s Safe to Retire Early (Part 2)

This is my second post in a two-part series presenting my early retirement calculations.  I’m providing this as an example that will (hopefully) help you decide when you’ve accumulated enough money for your own early retirement.  In Part 1 of this series, I argued that it’s probably unwise for most people to rely entirely on simple online retirement calculators and rules-of-thumb for this momentous decision.  Conducting calculations specific to your own situation is relatively easy, if you have decent high school math and spreadsheet skills, and will increase your confidence in your retirement decision.  My last post presented my own calculations using historical stock return scenarios, and I found that:

  • My retirement spending plan has an inflation-adjusted 2.8% annual rate of withdrawal.
  • My retirement portfolio succeeded (doesn’t run out of money) in all the scenarios that used an inflation rate ranging from 2% to 4%.

Here’s a graph showing the 14 scenarios using historical sequences of stock returns going back to 1871 and the 2% inflation assumption.  Assuming 2% inflation, my portfolio was not depleted in any of the scenarios by the time I would be 86 years old.

In this post, I’ll evaluate these results to illustrate how I made my final decision to go ahead and retire early in January of 2018.  If you haven’t read Part 1 of this series, I suggest you do that before proceeding further here.  If you understandably can’t stomach reading two of my posts in succession, you should be able to understand the major points of this post just based on the above results summary.

As I noted in my last post, these results raise some questions that deserve more attention.  Answering these questions, using my situation as an example, should give you a general template for evaluating your own early retirement calculations.

What’s the best way to include inflation?

My portfolio seems almost unsinkable using a 2% to 4% range of inflation.  I used nominal (not inflation adjusted) returns, and I increased the withdrawal rate annually by the assumed rate of inflation to maintain my standard of living as retirement proceeds.

However, another equally valid (some would say “better”) way of doing these projections would be on a real-return basis, which automatically incorporates historic sequences of fluctuating inflation rates.  But there a pros and cons to both nominal and real return projections.  The problem with the nominal approach is that stock returns are correlated, although erratically, with inflation rates.  So, some would say that using historical sequences of nominal returns but with a constant inflation rate ignores this historical correlation.  Conversely, the real return approach problematically assumes that future inflation sequences will be linked to future return sequences exactly like they were in the past.

Because the historical correlation between stock returns and inflation has been very erratic, I find it unrealistic to assume that future returns and inflation will move together like an exact copy of the past.  So, I prefer the nominal approach where various constant inflation rates can be used to see how the results might change when inflation fluctuates.  As the far right column in this table shows, inflation rates have varied considerably since 1871.

Scenario First Year Last Year Nominal Annualized Return (CAGR) Annualized Inflation Rate
1 1872 1905 7.2% -1.2%
2 1882 1915 6.3% -0.2%
3 1892 1925 7.9% 2.6%
4 1902 1935 7.2% 1.7%
5 1912 1945 8.0% 2.0%
6 1922 1955 11.0% 1.4%
7 1932 1965 13.5% 2.4%
8 1942 1975 11.5% 3.7%
9 1952 1985 10.9% 4.3%
10 1962 1995 11.3% 5.0%
11 1972 2005 11.0% 4.8%
12 1982 2015 11.3% 2.8%

Deflation prevailed in the late 19th and early 20th centuries, and in the latter half of the 20th century inflation greater than 4% was common.

But today, the Fed is projecting “longer-run” inflation at around 2%.  A 2% rate is also the Fed’s long-term goal for Goldilocks inflation that’s neither too hot nor too cold for a healthy economy.  It seems likely that the Fed would take extraordinary steps (like those of 2008-09) to avoid the extreme deflation/inflation cycles of the early 20th century or the runaway stagflation of the 1970s.  So, I’m pretty comfortable approximating this inflation dynamic using a reasonable range of constant inflation rates.

Of course the beauty of doing your own projections is that you can choose either the nominal or real-return methods or both.  You could even use a hybrid approach where you vary the annual inflation rate based on the distribution of past inflation rates, but not necessarily in strict correlation with nominal returns.

Obviously, there’s some combination of extreme inflation and/or low return assumptions that would make almost any retirement plan fail, which implies it’s never safe or anybody to retire.  For example, I could instead assume the nearly 10% annualized inflation that occurred from 1974 to 1984.  But personally, I’m not sufficiently concerned about this unlikely “disaster” scenario to make it the acid test for my retirement decision.  A mindful perspective helps show that:

  1. It’s generally unproductive to worry about unlikely disaster scenarios.
  2. If a disaster occurs, having a few more dollars saved up probably won’t help matters much.

How do my results compare to generic online calculators?

As noted above, if I boost the inflation assumption to 4%, the “worst start” scenario comes pretty close to failing.  Here’s the graph for the 4% inflation assumption.

The rule-of-thumb for a safe withdrawal rate is 4%, and my withdrawal rate is a comparatively low 2.8%.  So, it makes sense that none of the scenarios fails completely.  But given one scenario comes close to failing, it seems prudent to compare my results to an online retirement calculator.

FIRECalc is one of the most flexible online calculators that I’ve come across.  It requires three basic inputs: starting portfolio value, expected annual spending, and years of retirement.  It outputs the chances of portfolio success by looking at all sequences of historical returns going back to 1871.  FIRECalc also allows somewhat limited inputs for other income sources, variations in spending rates, inflation rates, and asset allocations.

When I enter values into FIRECalc matching my own calculations as closely as possible including 4% inflation, FIRECalc reports that my portfolio was successful (didn’t run out of money) in all 45 scenarios calculated.  The smallest final terminal value of any FIRECalc scenario is nearly identical to my “worst start” scenario.  But FIRECalc calculates a maximum final value for the best performing scenario that is well below the maximum final value from my calculations.

What’s causing some of these differences?  After a careful review of FIRECalc and its supporting documentation (which is pretty vague), I identified three potential factors that are difficult to control for:

  1. I have a some large one-time spending events in the first 10 years of my calculations beyond the base 2.8% withdrawal rate, which I could only roughly imitate in FIRECalc.
  2. FIRECalc assumes annual rebalancing, while I use the sequential depletion approach noted in my last post, which avoids rebalancing.
  3. The compounding frequency in FIRECalc is unclear, while my calculations use monthly compounding.

There may be some other differences as well, but these seem to be the main ones.  Based on the available documentation, it’s hard to say whether the first factor would drive FIRECalc results to be more or less optimistic than my own results.  The second factor of rebalancing could also drive results either way, because regular rebalancing doesn’t always improve returns.

The third factor of compounding could be causing some of the differences in the maximum final portfolio values between FIRECalc and my results.  Given that FIRECalc starts with Robert Shiller monthly stock return data, it seems very possible that FIRECalc uses monthly compounding like I did.  However, the description of the calculation steps provided in FIRECalc seem to imply that they use annual compounding.*

These differences show how the specifics of your situation and relatively trivial-sounding calculation methods can make a big difference to your results.  While FIRECalc and its kind are very helpful, they can rarely mimic your exact situation, and they are all opaque to some extent on the exact details of the calculations.

Could I adapt my spending to further increase my chances of success?

I’m comfortable with the idea that there’s an outside chance my retirement plan could fail if I experience a particularly bad combination of poor early returns and sustained high inflation in the next 34 years.  But the real clincher to my early retirement decision is the flexibility I have to reduce my spending, should disastrous scenarios like this start to play out.

There’s plenty of debate about whether real spending in retirement is more likely to stay constant (per the 4% rule), decrease, or increase at the start and end of retirement.  Regardless of the merits of each argument, there are surely retirees out there who have, for idiosyncratic reasons, experienced all these spending trajectories as they age.  Personally, I’m absolutely certain that I could forego all the one-time spending events I included in the first 10 years of my calculations because they’re mostly for optional purchases like extended vacations.

So, I re-ran the “worst start” scenario assuming 4% inflation but eliminating all those optional purchases in the first 10 years.  Here’s how the key result (ratio of the final portfolio value to the starting portfolio value) changes:

  • Full spending – 0.67
  • Eliminate optional spending – 3.18

This one change increases my final portfolio value by nearly 5 times.  If the stock market tanks early in my retirement, I can simply forgo the optional purchases I built into my other projections.  And if that’s not enough, I’m pretty comfortable with the idea of not giving myself a cost-of-living increase each and every year.  Alternatively, I could limit my increase to 4%, even if inflation exceeds that level for a year or two.  All of these adjustments will help my portfolio withstand a disaster scenario.

Even further, as I’ve presented before, I plan to use cash to buy more stocks in the event that a major stock crash (very specifically defined) occurs in the first 8 to 10 years of my retirement.  This once-in-a-lifetime market timing decision would further bolster the recovery of my portfolio in most cases.

Conclusion

The differences between the results of my homemade calculations versus an example online calculator cause me some concern.  Because the exact methods of online calculators are rarely entirely transparent, it’s impossible to ferret out all the reasons for the differences.  All these calculations attempt to predict a highly uncertain future, so there’s no point in trying to find the one “right answer”.  Rather, the different results from similar inputs are best viewed as a measure of the margin of error associated with the decision to retire early.

When I step back and review the entirety of this exercise, it paints a picture of a very safe retirement plan.  If inflation remains within about the 2 to 4% range, my retirement plan will likely be successful by almost any measure.  If inflation really takes off in the next 30 odd years, and the stock market performs poorly at the same time, then I may have a bumpy ride.  But even in this case, I’ve got three layers of counter measures at the ready to bolster my portfolio’s longevity.  So, that’s why I decided at the start of this year that it’s reasonable to retire at the age of 52, but only so long as I keep a careful eye on my portfolio, inflation, my spending, and how these factors play out over time.


*I used monthly compounding because the Shiller data are set up in the form of monthly returns.  However, I find the monthly compounding assumption (also used by some other calculators and studies) somewhat unrealistic.  All these stock return sequences use the price and dividend history of the S&P 500.  Stock price changes occur daily and withdrawals can reasonably occur on a monthly basis.  But dividends from the S&P 500 are paid no more often than quarterly, regardless of whether you invest using exchange-traded funds, mutual funds, or individual stocks.  And reinvested dividends represent most of the wealth generated by the S&P 500 since 1871.  This may be the reason my maximum final portfolio values exceed those from FIRECalc for the higher-performing scenarios.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.