Hacking The Retirement Calculators

23Investing, Money, , ,

How was your first date?

My first date movie experience? As memorable as this.

Using a retirement calculator for the first time is somewhat like going out on a first date.  

It is a rite of passage on your journey towards financial independence, much like the first date is a prelude to an enduring relationship.  It explores sides of you that didn’t know existed! Were you a clumsy, bumbling teenager who wanted the first date to be over with really fast?  

I was happy to get the dinner over with (the pizza was yummy but the company was not).  I painfully sulked in my seat as some terrible movie (not my choice, obviously) was on, while slowly munching from a large bowl of popcorn.  I don’t even remember the movie name now.

Like the first retirement calculator.

Retirement Calculators – Can’t live with them or without them

That experience is somewhat akin to my first time use of an internet retirement calculator, probably the first ever retirement calculator on the net.  

I think it was at a CNBC website right around the time such calculators were born and had barely 4 data fields before it spit out the answer.  I am talking about the pre-dot com era, long before sophisticated retirement calculators based on Monte Carlo simulation or back-tested real data started appearing.  

The comparison is like an IBM monochrome PC to today’s super fast graphics chip powered loaded laptop.  

Much like a clumsy teenager, I was quick to try it and when faced with a result I didn’t like, I forgot about it and didn’t call on the site ever again.  That was when my net worth was in 4 figures. I remember being depressed about the result that the retirement calculator spit out.

Like today’s retirement calculators. Hack-worthy.

Today’s calculators are a class apart.  While there are some good calculators on fund management sites (Vanguard, Fidelity, T. Rowe Price websites for example), there are two that particularly stand out for the aspiring financially independent or early retirement (FIRE) community.  

These calculators are particularly good from the perspective of comprehensive historical database, ease of use and wide input flexibility for scenario analysis.  I am specifically referring to FIRECalc and cFIREsim.

Still, we should remember these calculators rely on historical performance to make projections.  Keeping the adage past performance is not predictive of future results, how do we coax valuable information out of them, particularly at the time of historically low bond yields?  

Most people do the analysis over the past 30, 40 or even 50 year horizon for a conventional 60/40 portfolio of stocks and bonds. In that entire time period nowhere have the bond yields been this low as we are seeing now.  Some say we have experienced a generational bull market in bonds that is finally showing signs of cracking.

Hack your way to more money in retirement.  

In this backdrop, how do we reliably use the retirement calculators? By hacking them, of course!  

Hack the calculator to your advantage.

Relax with a nice cuppa as we look into this simple but incredibly effective way to calm your retirement nerves in current times.

The first hack we will do is to eliminate bonds from analysis.  By defaulting to a 100% equity portfolio analysis, we eliminate the entire question surrounding the historically unique bond market today.  So, how do we correct for the fact that many portfolios have bonds?

Well, we don’t!

This is where Hack #2 comes in.  By demanding a very high probability of success, we force the calculator to come up with the lowest possible spending rate. This takes care of the bond allocation in these low yield times (up to 30% bond allocation, based on my analysis.  A 80/20 portfolio behaves practically the same as 100% stock).  

Since the current stock valuation is still within the range of historical P/E ratios, by demanding a very high (98-100%) probability of success, you force the calculator to become conservative. This serves as a proxy for the relatively high stock market valuations.

Hack #3 is to ignore Social Security or even pensions (unless you are absolutely sure of getting it).  Retirement calculators consider these ‘guaranteed’ income streams as direct offset against withdrawals, so the moment you enter a pension or SS, your safe withdrawal amount shoots up considerably.

If you have a fixed withdrawal target, your probability of success catapults much higher.  By eliminating this factor, you are making an indirect allowance for a high P/E ratio of current markets in the calculations and getting to the truly safe withdrawal rate from your portfolio.

Let’s see an example.  We will calculate the safe withdrawal rate (SWR) for a 50 year retirement horizon that gives 100% chance of success for a $1 million, 100% equity index portfolio invested at a very low 0.1% annual expense ratio with the assumption of no social security or pension.

You can run the simulation here using the above settings, Run Simulation with the Investigate menu set at ‘max. initial spending’.  

With these hacks, cFIREsim reports $32,752 as the maximum initial spending possible.   This works to a 3.27% withdrawal rate in the first year of retirement.  My fellow blogger Early Retirement Now has published a large series of posts analyzing the SWR under various scenarios.  

This simple hack we did brought us to essentially the same conclusion that ERN arrived at after massive amount of analytical work (which is commendable, and worthy of a university thesis).

What does this mean?

We should remember that we did these hacks with the sole purpose of finding the edge of the failure scenario.  Put another way, this gives you a truly ‘safe’ withdrawal rate for the first year of retirement, assuming the worst historical case scenario, as a conservative proxy for current relatively high stock valuations.  

But that’s all you should treat these results as – don’t read too much into them.  Don’t get enamored with the median or max portfolio values from this simulation – you will notice that they are very (possibly, unrealistically) high.  

This is because they are an artifact of the 100% equity portfolio assumption we deliberately made to capture the widest possible swing in portfolio values.  The widest swing naturally includes a very low point (at the verge of failure) and a very high point (max value) and several higher value ranges (median and above) of market performance, none of which would apply if your portfolio has bonds or even stocks of different types (sector focus or dividend-focus) than the index.  

Even if you don’t hold any bonds, I would treat the higher portfolio values with a big pinch of salt because this simulation doesn’t correct for current stock valuation.  Therefore, the most useful result from these hacks is the ‘lowest point’ (or failure edge) of SWR, which is converged by the 100% success probability we demanded at the start.  

That SWR is 3.27% – which is where you can start your first year withdrawals at. Notice how this is considerably lower (18.25% lower, to be precise) than the famous 4% rule.

With this ‘safest’ case now done, you could relax the success probabilities down and calculate SWR (keeping all other variables the same as above).  Doing that gives me these results below (for the same example):

SWR reduces with increasing probability of success.  I know, right?

We saw in an earlier article how the probability of success in a retirement calculator is not a guaranteed way to retirement success, given the complex reality of our lives.  

Even Burton Malkiel, the author of the famous book, A Random Walk Down Wall Street, says that once you cross 80% probability of success in such simulations, you are practically good to go because there are so many life variables involved that determine your ultimate success. So, going from 90% to 98% gives a false sense of assurance when the other risk factors become more significant.  

With this caveat in mind, see how the SWR varies with the ‘required’ probability of success – a difference of 34% in income from the 100% to 80% case.   Even the 14.3% difference between 95% and 100% case is meaningfully large from a spending perspective.  That’s almost $5,000 in additional spendable income.

The same table can be interpreted in another way as well.   You can also see why any kind of a small, part-time income, or even a modest pension or social security, dramatically improves your probability of success.  

If you need $40K to live on, a small side income of $7,500/year or a tiny pension/social security check of $600/month can catapult your portfolio’s success probability from 90% to 100%.  

From a one-in-ten chance of failure to a near zero chance of failure! So, don’t get scared by the lower probability of success figures because you are lot more in control of your life than what these retirement calculators assume.

To sum it up…

An objective of this exercise is to hack the retirement calculators to counteract the current scenario of high stock valuations and low bond yields to arrive at a truly ‘safe’ withdrawal rate to ensure your retirement success.  

We have achieved that objective and arrived at an SWR that essentially matches a much more complex, analytically rigorous study.  Using these simple hacks, we have established that 3.27% as our new ‘safe’ (safest?) withdrawal rate by converting a quirk of a rigorous retirement calculator (that is, assign probability of failure to the ‘edge’ event) to our advantage.  

In reality, what’s safe is not a single SWR number – from the above analysis, we can consider the safe range as 3.27%-3.74% because a lot depends on how the portfolio evolves during the first decade of retirement.  But remember, the results from the hacks aren’t valid for any other purpose if your actual portfolio has a different allocation.  

What do you think?  Are you less nervous now about your SWR?  Share your feedback below.

Added later:  Based on some comments this article received, I think it is useful to add a caution that a retirement calculator based on historical data used with sufficiently conservative assumptions is the closest you will get to estimating your retirement success.  Since the market is based on secular trends and is not random (see my reply below in response to a comment on this topic), I would urge you so stay away from Monte Carlo-based simulation.   To read an excellent paper on why Monte Carlo simulators aren’t the best tool for retirement planning, see this article by Jim Otar.

Like it? Share it!

23 comments on “Hacking The Retirement Calculators”

  1. By The Money Commando

    I love the analysis, and your conclusion of a roughly 3.25% SWR matches what I’ve come up with using a variety of other methodologies. The part I struggle with (and you both mentioned and tried to address) is adjusting for the current high market valuation.

    I’m personally planning on a 3% withdrawal rate, with the idea that it’s much easier to adjust spending upwards than downwards!
    The Money Commando recently posted…Investment income – September, 2017My Profile

    • By TFRadmin

      Thanks TMC. You are way more conservative than needed in my view. We have had high stock valuations like these in the past and combined with stagflation type scenarios. Being in 100% stocks starting retirement in 1966 is the historical worst case scenario over the past 70 years in many ways. I don’t believe today’s is even close by comparison. The 3.27% is based on the worst historical case (failure edge) so I don’t believe in going lower than this SWR. Unless you feel a Great Depression is likely but even then, there was deflation for a decade after which helps in lower spending. This wasn’t the case for 1966 retiree because in 70’s severe stock declines was accompanied by high inflation rates – a decade of pure misery for retirees. 3.27% SWR has survived that!

  2. By 401k Manager

    As a financial planner, I totally understand where you’re coming
    from. I read your site fairly often and I enjoy your posts.

    I shared this on twitter and my followers enjoyed it too.
    Kepp up the good work!
    401k Manager recently posted…401k ManagerMy Profile

  3. By Make Wealth Simple

    Great piece. And a fun new blog to add to my reading list.
    I’m saving this post for reference and will look in future how I can apply these to calculations for a European context ( where financial markets have been less dynamic than the U.S. market, and taxes are set up a bit different). The SWR of 4% was not a good number for Europeans I think, but it is more easily marketable than 3,27% 🙂 (Although I definitely think that you are closer to the truth).

    • By TFRadmin

      Thanks for stopping by MWS! I am glad you liked it and happy to have you on this site. You can always adjust taxes accordingly as all withdrawals are calculated Pre-tax (tax is an expense like any other) and Yes, 4% is lot more marketable than 3.27%!

  4. By Amber tree

    Thx for analyzing this. It helps peopel to better undersatnd the pitfalls of teh SWR number.

    Being far away from FI, I consider a 4 pct SWR. As you say, there are soo many variables in life, so many things that can happen: take a small side gig, be flexible on spending,…

    In reality, I see FI as means to an end, not the end goal. I do expect to have some years with a SR close to0 as it will be a year with a lot of upaid leave and far and adventurous travel.

    That makes that I use 4pct to track my progress to FI. I think it is important to see progress i my FI efforts.

    • By TFRadmin

      Thanks ATL. It is important to be comfortable with the SWR number knowing you have so many variables under your control even if the future is very different than even the ‘worse case past’. Human nature being what it is, we overplay disaster scenario (which is very rare) in our planning rather than even mediocre outcomes (which are more common) and extremely optimistic outcomes (also rare). This approach helps to a degree but beyond that it can cause ‘analysis paralysis’.

  5. By Physician on FIRE

    I plan on doing something similar with our withdrawal rate. Take what’s assumed to be pretty safe, and spend less than that. I anticipate spending in the 3% to 3.33% range.

    Better safe than sorry!
    -PoF

    • By TFRadmin

      Thanks PoF. With an approach like that, you are on your way to the Forbes list!

  6. By FullTimeFinance

    I’d actually recommend running vanguard or one of the other simulators and one of the two you mentioned. I have a post on this in my featured post menu of my blog. That being said fire call and cfiresim use similar methodologies. There are two popular methodologies to retirement calculators. One uses montecarlo simulations on individual yearly returns. The other you choose a year range, say thirty years. It then determines what would happen for every one of those year ranges. This is what fire alarm and cfiresim do. There are mathematical limitations to both approaches so I find it best to mix and match.

    • By TFRadmin

      FTF, Thanks for your comment. This article isn’t about which retirement calculator is good – there are many. It is specifically about using a distinct feature of the historical data-based calculator to get the data we want. It is about assigning probability to the ‘failure edge’ event to arrive at ‘really safe’ withdrawal rate in current market conditions. I trust long series historical data-based calculators more than Monte Carlo simulator because the latter is purely mathematical and has no basis in investor psychology, whereas historical data based projections have built-in investor reactions that caused market movements at those times. While history may not repeat itself, the underlying human psychology always does, so it is much more reliable as a guide. Unless you believe that the failure edge event (say a 1966 retiree facing nearly 15 years of down market conditions) is not stringent enough for the future. We have to draw the line somewhere to sleep well at night, so this method, I believe, achieves that objective to achieve a really safe SWR. It is probably overly conservative but still, if a retiree can’t sleep well at night at 3.3% SWR, the problem isn’t money but something else.

      • By FullTimeFinance

        I like to look at both. Japan demonstrated the US history is not sufficient, but montecarlo misses out on the momentum effect. It’s a case of the more data points the better imho. That was really my point, both have flaws, but they are different ones. An aggregate of the different methods decrease you chance of failure.

        • By TFRadmin

          I understand FTF but I still prefer historical data-based simulations. For example, on Vanguard’s retirement calculator that uses Monte Carlo, the following fineprint appears as a weakness, which is the same point about investor behavior/psychology I mentioned in my previous comment:

          “It’s also important to remember that, despite the sophistication of this method, this calculator makes a number of simplifying assumptions, so these results should never form the sole basis of your financial plan. In particular, the Monte Carlo methodology used here assumes no relationship between asset-class returns from one year to the next. Randomly selected years are considered in sequence. For example, in a given simulation the returns on stocks, bonds, and short-term reserves for 1982, when the nation was deep in recession, could be followed by the returns for 1999, a bull-market year.”

          By Vanguard’s own admission, a 1999 market year appearing immediately after 1982 would’ve been impossible in real life but Monte Carlo includes it as possible.

          So, in my previous comment example of a 1966 retiree who faced a 15 year market ‘drought’ till 1982 is a real-life scenario that Monte Carlo would not consider. Most SWR failure cases involve 1966 retiree since, historically, that and also, a 1999 retiree were the worst times to begin retirement – both cases being stock-heavy portfolios. Actually, 1999 retiree is still faring a little better than 1966 retiree from the data I have seen. In historical data based retirement calculators, the toughest case (‘failure edge’) is 1966 retiree. If the simulation passes that, then you have a near 100% probability of success. In reality, living for 15 years under declining markets of which nearly a decade under the miseries of high inflation, no growth and high unemployment (‘stagflation’) with a stock-heavy portfolio would put retirement plans in serious trouble. If not for social security and banks paying high interest, most people would’ve been in abject poverty during those times. In order to benefit from the biggest equity bull market (1983-2000) of the century, a 1966 retiree must be very frugal in withdrawing from the portfolio for the first 15 years. That’s why this scenarios is the real stress test for most calculators and gives you the ‘failure edge’ (leading to SWR of 3.27%) as mentioned in my article. Monte Carlo can ‘dilute’ this scenario by bringing in a boom year like 1999 anywhere in the 1966-1982 sequence because its approach is totally random. Doing so distorts reality and gives a false comfort for the retiree facing a 1966-type start in real life.

          • By FullTimeFinance

            This is what I referred to as the momentum effect. The thing is, it’s not necessarily overly optimistic. Montecarlo does thousands of simulations. Improbable negative returns are just as likely as improbable positive. Over 5000 simulations you would have a bit of both, but on the whole it should follow the distribution shape of returns over the course of your sample period. I.e. You’d be more likely to see the average return of any given year in most of the 5000 line results. However the important part is how you view the data. It gives you a percent of results. Thus you can account for both extremes if the number is sufficiently high. It’s a statistically sound methodology, that has a higher likelihood of presenting a lower likelihood of success then firecalc. I.e. If one of 5000 is improbably negative you will get less the 100 percent results regardless of the number of overly positive results.

          • By FullTimeFinance

            Just to ensure it’s clear, I’m not saying it’s superior, I’m saying it’s making different assumptions. A data set of 150 periods as in firecalc of a country in a growth phase is not indicative of all scenarios. But it takes into account momentum, which is a great add. Sometimes you need to look at multiple methodologies to discount the risk of each methods assumptions.

        • By Ten Factorial Rocks

          I agree Monte Carlo is statistically rigorous but not more accurate. The inference isnt nearly as valuable. A lot of people talk about Japan but ignore the inter-generational peak they sat on 1989 against a poor demographic time bomb, combination of which has killed their stock market for 25+ years. Japan is a homogenous society that didn’t allow either immigration to refresh the demographic profile towards growth nor did they consider that efficiency alone is not a sustainable competitive advantage. Whether country or company, innovation, scale and diversity are critical for long term advantage.

  7. By The Green Swan

    Interesting analysis, 10! I always love giving articles like this a read. And good point about what Malkiel said.

    I like the way you went about setting an edge, it helps having that anchor in mind. And like you said, it’s a good starting point for that first important decade. But I’d love to hear what you think about that next decade. Once you make it through the first one, the next one becomes most important and you still have a long retirement ahead of you. Should your mindset change at all?

    Thanks for the great analysis!

    • By TFRadmin

      Thanks for your kind words GS. I believe in achieving a balance between analysis and action so we don’t get into ‘analysis paralysis’. 🙂

    • By Ten Factorial Rocks

      Also, GS, to your question, if you start the second decade with same amount of assets as you had a decade ago, despite a decade of withdrawals, then you’re good to go on using the same SWR inflation adjusted. I would suggest a ratio check every 5 years to make necessary course corrections.

      • By The Green Swan

        My goal would be to die with the same inflation adjusted value of retirement assets as what I started with. It’s conservative but necessary given the ultra long retirement horizon. That makes sense to do a ratio check frequently. Thanks 10!

        • By TFRadmin

          GS, it is an aggressive goal but having it will force you to go to the lowest SWR possible. But in life, there is a serious compromise. Perhaps you can can consider a happy medium of having the same Nominal value of retirement assets at the end, instead of inflation-adjusted value? This will ensure sufficient assets remain for legacy purposes and yet give you more spendable income while you are retired. It is still more conservative than many simulators that allow down to $0 in final asset value.

          • By Steve Poling

            This is where your observation about keeping a small side hustle going, or a SS check will have an inordinate positive effect on your probability of success. Early retirement, when you’ll see the greatest sequence-of-returns risk is precisely the time when you’ve the most flexibility. Late retirement will bring health issues that’ll make it much harder to boost income or cut costs.

            This is why I’m favoring a 100% equities position with a big cash hedge (for capital expenses) and rental income to supplement investment income. Add to that some modest side hustles and I should keep the withdrawal rate well below 3%. I hope that if I can live off the dividends VTSAX pays, I can forgo bonds altogether.

          • By TFRadmin

            Absolutely Steve. At just 2% div. yield, even a $40K/year from VTSAX dividends will require $2 million in capital. Not an easy goal for many but I am impressed that you have this goal. I agree that bonds have no place then. I also don’t believe in holding bonds if you have a well-structured dividend growth portfolio. Thanks for commenting.

Comments are closed.