Cameron Passmore CIM, FMA, FCSI

Portfolio Manager

Benjamin Felix MBA, CFA, CFP

Associate Portfolio Manager
  • T613.237.5544 x 313
  • 1.800.230.5544
  • F613.237.5949
  • 265 Carling Avenue,
    8th Floor,
  • Ottawa, Ontario K1S 2E1

FIRE? Here’s Why The 4% Spending Rule Does Not Apply to You

June 27, 2018 - 2 comments

The 4% rule, or the 25x spending rule (1/25  = 4%) are often cited as safe spending rules for retirees. The 4% rule originated in William Bengen’s October 1994 study, published in the Journal of Financial Planning. The spending rule involves taking 4% of your retirement portfolio as your first-year retirement income, and then adjusting for inflation each year going forward.

Bengen wanted to find the maximum sustainable withdrawal rate during worst 30-year period in US history. Using data from Ibbotson Associates Stocks, Bonds, Bills, and Inflation yearbook he constructed a hypothetical portfolio consisting of 50% S&P 500 index (US stocks) and 50% intermediate-term US government bonds. He then looked at each consecutive 30-year period starting in 1926, ending with 1992 (1926 – 1955, followed by 1927 – 1956 etc., ending with 1963 – 1992).The safe maximum withdrawal rate ended up being just over 4%. From this simple but innovative analysis, the 4% rule was born.

Another study was conducted in 1998 along similar lines but with slightly different data by three professors at Trinity University in Texas. This study showed that a 4% withdrawal rate has historically had a 95% chance of not prematurely depleting capital. The result of this corroborating study further increased the adoption of the 4% rule in financial planning circles and resulted in the commonly held belief that a 4% starting withdrawal rate is a safe approach for retirees.

More recently Bengen has adjusted his spending rule to 4.5% based on the inclusion of small cap stocks in the hypothetical historical portfolio. While the 4% (and the 4.5% rule) may have basis in historical US data, there are substantial problems with these rules in general, and specifically in the case of a retirement period longer than 30 years.

Non-US Returns

Betting on history repeating itself is inherently risky, but betting on the performance history of a single country is even riskier. Most sensible investors will invest in a globally diversified portfolio. In his 2017 book How Much Can I Spend in Retirement, Wade Pfau, Ph.D, CFA, looked at 30-year safe withdrawal rates in both US and non-US markets using the Dimson-Marsh-Staunton Global Returns Dataset, and assuming a portfolio of 50% stocks and 50% bills. He found that the US (3.9%), Canada (4.0%), New Zealand (3.8%), and Denmark (3.7%) were the only countries in the dataset that would have historically supported something close to the 4% rule. The aggregate global portfolio of stocks and bills had a 30-year safe withdrawal rate of 3.5%. As previously mentioned, betting on history repeating itself is risky; there is substantial survivorship bias built into the data. In 1900, we would not have known which markets would continue to exist for inclusion in the data set.

Historical Retirement Periods Longer Than 30 Years

In Bengen’s study the 4% rule with a 50% stock 50% bond portfolio was shown to have a 0% chance of failure over 30-year historical periods in the US. That chance of failure increases to around 15% over 40-year periods, and closer to 30% over 50-year periods. FIRE likely means a retirement period longer than 30 years; looking at longer time periods is likely prudent for an early retiree. One issue with using historical data is that we are limited in the number of samples that we can take, especially for longer sample periods. One way to address this issue is with Monte Carlo simulation using reasonable expectations for future returns.

Expected Future Returns

There is an obvious risk to using historical data to build expectations about the future. It may be reasonable to expect relative outcomes to persist (stocks outperforming bonds, small stocks outperforming large stocks, and value stocks outperforming growth stocks) but the magnitude of future returns are unknown and unknowable. The world today is different than it was in the past; interest rates are relatively low, and stock prices are relatively high. Building out a set of expected returns for financial planning is neither an art nor a science but it is a necessity in making many financial decisions. PWL Capital uses a combination of equilibrium cost of capital and current market conditions to build an estimate for expected future returns, as outlined in the 2016 paper Great Expectations. With reasonable expected return and standard deviation assumptions as inputs, we can use Monte Carlo simulation to test safe withdrawal rates. Using expected future returns as opposed to past returns may give us a more reasonable estimate.

Monte Carlo Safe Withdrawal Rates

Using the December 2017 PWL Capital expected returns for a 50% stock 50%i bond portfolio we are able to model the safe withdrawal rate for varying durations of retirement using Monte Carlo simulation. We will initially ignore taxes as that is how the 4% rule was originally presented. We will assume that a 95% success rate over 1,000 trials is sufficient to be called a safe withdrawal rate. 

To speak to the FIRE philosophy, we have pushed the retirement period as far as 55 years to model someone living off of their assets from age 40 to age 95 with no other sources of income. We assumed constant 2% inflation.

1,000 Simulations Ignoring Taxes

Years of Retirement Safe Withdrawal Rate (1st Year)
30 3.5%
35 3.1%
40 2.9%
45 2.6%
50 2.4%
55 2.2%


Factoring in taxes is an important part of determining a safe withdrawal rate. Taxes are a reality and a constant. The interesting thing about modelling taxes and withdrawal rates is that the size of the portfolio starts to matter. On a $1M portfolio there is little taxable income above the basic personal exemption for someone with no other income sources. The result is only slightly lower after-tax safe withdrawal rates compared to a tax-free investment.

1,000 Simulations w/ Taxes on a $1M Portfolio

Years of Retirement Safe Withdrawal Rate (1st Year)
30 3.4%
35 3.0%
40 2.8%
45 2.5%
50 2.3%
55 2.1%


Increasing the portfolio to $10M will correspondingly increase taxes, decreasing the safe (after-tax) withdrawal rate for spending.

1,000 Simulations w/ Taxes on a $10M Portfolio

Years of Retirement Safe Withdrawal Rate (1st Year)
30 3.1%
35 2.7%
40 2.3%
45 2.1%
50 1.9%
55 1.8%


Flexibility in Spending

One way to increase the amount that you can spend overall is allowing for variable spending. There are many different variable spending strategies that can be applied to a portfolio of risky assets. The result is more spending overall with a lower probability of running out of money. The catch is that you have to live with a variable income or have the ability to generate income to fill in the gaps when markets are not doing well.


All of the analysis mentioned so far has ignored fees. Fees reduce returns. In his book, Pfau explains that in the historical US data, adding in a 1% fee to the 4% rule analysis reduces the safe withdrawal rate anywhere from 0.48% to 0.86%, with an average of 0.65%. The safe withdrawal rate in the worst 30-year period in the US drops to 3.56%, making the 4% rule the more like the 3.5% rule after fees. Adding a 1% fee to the Monte Carlo simulation with future expected returns reduces the safe withdrawal rates by around 0.50% on average. In both cases this is a meaningful reduction in spending. Of course, fees need to be considered alongside the value being received for the fee. This value is heavily tied to behaviour and financial decision making. 

Financial Advice

There have been two well-known attempts to quantify the value of financial advice, one by Vanguard and one by Morningstar. Vanguard estimated that between building a customized investment plan, minimizing risks and tax impacts, and behavioural coaching, good financial advice can add an average of 3% per year to returns. Morningstar looked at withdrawal strategies, asset allocation, tax efficiency, liability relative optimization, annuity allocation, and timing of social security (CPP in Canada), to arrive at a value-add of 2.34% per year. PWL Capital’s Raymond Kerzérho has also written on the topic, finding a value-add of just over 3%. Based on these analyses, one could argue that paying 1% for good advice could even increase your safe withdrawal rate. I would not go that far, but the point is that while fees are a consideration they may be worthwhile in exchange for good advice. Of course, if you can stay disciplined, take the time to educate yourself on financial planning, keep up-to-date with tax laws and investment products, and maintain your cognitive abilities into old age, you may be able to make your own good financial decisions without paying a fee for advice.

The 4% rule is a constant inflation adjusted spending strategy. It may have worked over the worst 30-year period in US history, but confidence in the rule starts to decrease when we consider international data, longer retirement periods, and expected future returns. A more likely safe withdrawal rate in a 55-year FIRE situation might be closer to 2% before fees are taken into account. Fees are a consideration in determining a safe withdrawal rate; all else equal, lower fees mean a higher withdrawal rate. However, there are many financial decisions along the way that could impact sustainable lifetime spending. Despite adding a cost, financial advice may end up being beneficial over the long-term.


4.52% total return w/ 6.18% standard deviation composed of 1.80% interest and foreign dividends, 0.40% Canadian dividends, 1.16% realized capital gains, and 1.16% unrealized capital gains.

By: Ben Felix with 2 comments.
  05/07/2018 2:07:02 PM
Ben Felix
Grant, thanks for the question. We have only a relatively small handful of periods in the past where the CAPE at the start of a 30+ year period was as high as it is today. Assuming that those few periods are sufficient to plan for the future would be naïve. We do not know what sequence of returns will follow today, but we do know that it is unlikely that the sequence will match that of the past. By adjusting expected returns for current valuations and running 1,000 simulations that do not rely on the historical data we get a more reasonable estimate of a safe withdrawal rate. The blog that you linked is great but they are still using historical US data which is not necessarily descriptive of the future and suffers from substantial survivorship bias.
  02/07/2018 5:19:16 PM
Ben, very interesting post. The Bengen studies (like this link below) use just historical returns because included in this returns is a time period (1929) when the CAPE was even higher than it is now. So I'm wondering why you elect to include today's high valuations in the input data? Isn't today's high valuation already baked into the historical data as one of the possible outcomes as it has occurred before. So why not just use historical data? There is quite a difference in withdrawal rates in this article, although using the US as one's home country would also help to increase the withdrawal rate due to higher US returns over the last century.

 Security code