Monte Carlo Simulation: A Powerful, but Imperfect Tool

Working with financial planning clients over the years I have heard a wide range of individual goals, from providing college education and weddings for children, to purchasing vacation homes and sports cars. Every person has their own unique wants and needs shaped by their values, beliefs and attitudes. However, one goal every single person has shared in common is retirement.

Even with that common goal of retirement, the vision for that word is unique to everyone. One thing is clear though, we have entered into an age of unprecedented longevity, and with modern medicine that trend is only set to continue. How do we prepare our finances for a “retirement”, that may last 30, 40, 50 years or even longer?

Retirement is a unique goal for a few reasons. First and foremost, the income dynamic switches. For most people this means changing the mindset from saving money to spending the savings. It can be difficult to shift mindsets with the flip of a switch and feel at ease spending the money you worked so hard to save. Second, we don’t know when we’re going to die. If you tell me you’re going to die at exactly age 90 and don’t want to leave a legacy to heirs, that’s fine; we can spend down your assets so you pass away with nothing to your name.

For most people this isn’t the case. The thought of outliving our money and going broke in retirement is a very real fear for a lot of people. It’s why the data shows people have a really hard time actually spending down their nest egg, especially among higher net worth individuals with $500k or more of investable assets. This is why the two questions we hear most often from clients and prospective clients alike are, “how much do I need to save for retirement?” and “how much can I spend in retirement without running out of money?”

Thankfully, as with most industries, technology has vastly improved the field of financial planning and in-turn brought more science into retirement planning through statistical modeling. Monte Carlo simulation uses market data, either historical or projected, to predict outcomes through thousands of individual simulations based on investor specific inputs (variables). In just a few seconds, a “Probability of Success” for retirement is generated. A score of “0%” means there is no chance of meeting your retirement goal. A score of “100” means there is absolute certainty you will meet your retirement goal. Most people fall somewhere in between these two probability of success outcomes. I will explain more on the Probability of Success further along in the blog, but for now, remember the program is like any technology and only as strong as the data inputs that are entered into the software.

Monte Carlo simulation has been a vital tool for financial planners and clients alike by providing statistical data to support retirement spending habits and assuage fears about running out of money. But it has also been a misused and abused tool by financial advisors, leaving clients with a muddled understanding of where they stand in relation to their retirement goal(s), both now and in the future. Monte Carlo is also making predictions about the unknown via simulation of outcomes. We don’t know what the future holds for financial market returns and what life might throw our way on an individual level. In this month’s blog, we’re going to review the strengths and weaknesses of Monte Carlo simulation, how the Probability of Success metrics should be viewed and used, and some final food for thought.

Strengths of Monte Carlo Simulation

Simulations and Building in Uncertainty: Monte Carlo is capable of running a thousand (or more) simulations with the click of a mouse. It would take days or even years to run these simulations long form. This saves advisors and clients alike time and money when conducting retirement analyses.The program is also able to incorporate uncertainty into simulations. Uncertainty and adverse market conditions are something we as investors know all too well. Recessions, depressions and markets in turmoil have happened in the past and are bound to happen again in the future. While it is important to have long-term average risk and return assumptions for projecting median outcomes in retirement, it is also important to see how adverse financial market conditions affect investment portfolios and meeting retirement goals. According to Hartford Funds, “there has been a bear market on average every 3.5 years” going back to 1929. Wars, famines, pandemics, and natural disasters have all happened globally within the last century. Monte Carlo provides advisors and clients perspective on what can happen if things don’t go according to plan and divert from historical averages. More to come on this later.

Engagement & Understanding: From a financial planner perspective, the interactive capabilities of modern Monte Carlo simulation software have been very beneficial for helping clients understand how various decisions directly affect the probability of success of their plan. One of my favorite tools is called the “scenario analysis” module. The module enables people to visualize how decisions directly affect outcomes. Clients and advisors together are able to compare side-by-side multiple scenarios, adjusting variables, to fit the needs of the client(s). Now, compare the scenario analysis module with your financial advisor handing you a 100 page PDF print out of a financial plan report. Where are my kinesthetic and visual learners at? Which would you choose? It is night and day for me; give me the hands-on interactive experience.

Levers: Imagine doing a problem and each time you want to adjust a variable you have to start all over. Monte Carlo simulation allows independent inputs to be adjusted one at a time, or multiple all at once. The retirement scenarios analysis typically starts with analyzing the “current” scenario. Often people have a success rate that is not adequate to meet their retirement goals. From here, adjustments can be made to variables as needed: retirement age, savings rate, retirement expenses, and investment returns (just to name a few), can all be adjusted. I like to refer to the variables as “levers” that can be pulled. For one, most of the inputs in the module have sliding scales that can be adjusted and look like levers. Two, I think of adjusting the variables like a lever; you can throttle it up to increase or throttle it back to decrease.

Encouragement: For some, it may only take changing one input such savings rate to increase successful outcomes. This may just mean maxing out your 401(k). For others that are further off track for retirement it may mean adjusting multiple inputs, such as delaying retirement age and decreasing retirement expenses. Everyone has personal preferences on which levers they would like to pull; there is no right answer. Oftentimes people get discouraged when they see a low probability of success. Monte Carlo simulation have been empowering for many people, showing that small adjustments to habits can bring about big changes in meeting long-term goals.

Weakness of Monte Carlo Simulation

Accuracy: If you tell me you’re going to save $2,000 a month and you are only putting aside $200, we’ve got a problem. If you put together a budget that tells me you’re spending $5,000 a month, but you left out all of that fun discretionary stuff like shopping and you’re really spending $8,000 a month, we got a problem. The program and its outputs are only as accurate as the data that is input. Yes, it is boring and can be time consuming to round up documents and put together a spreadsheet or two, but it is vital for creating an accurate plan. I once witnessed a senior advisor at a firm tell his associates to “guess” on client data inputs when the client did not provide information. And it wasn’t just one or two pieces of data, sometimes it was half of the data inputs. In the sense that information is inaccurate, whether willfully or on accident, the Monte Carlo simulation will not be accurate. If you work with a financial planner, it is essential that you provide honest and up-to-date information when using financial planning software.

Assumptions: Whereas the individual must provide accurate data respective to their personal situation, the advisor must provide accurate assumptions pertaining to risk, returns, inflation, taxes, and more.

  • Risk & Return: Some advisors like to use historical data and some like to use projected data. I don’t think there is a wrong choice here; to each their own, as long as the data assumptions are realistic and you are consistent with your method. Flip flopping between historical and projected returns to maximize asset returns is only going to hurt your plan accuracy by providing unrealistic return assumptions. The two methodologies will even out in the long term. For example, when stock valuations were high in 2021, forward projected returns were lower than historical averages. When the market sold off in 2022, some forward projected returns were higher than historical averages. Pick a risk return method and stick with it! Unlike client inputs (Levers), manipulating asset returns to improve the probability of success is not the way to go.
  • Inflation: We hear the term inflation thrown around a lot (especially over the last 24 months) and it’s generally used in a singular term to describe CPI (consumer price index). Historical inflation figures have varied widely across different categories. While the domestic average annual inflation measure by CPI has generally been lower since the 1980’s, the same cannot be said for things such as Health Care and Education, both of which have had “long-term” inflation rates north of 5%. These two categories also tend to have outsized impact on our lives and thus financial plans, especially health care costs for retirees. The Wall Street Journal reported last week that “Employers and workers are expected to see an increase of about 6.5% or higher in health-plan costs next year.” We tend to err on the conservative side while making these assumptions. Inflation coming in at just 1% higher than projected can have outsized negative effects on the outcome of your retirement.
  • Taxes: Death, taxes, and you fill in the blank with your third guaranteed in life. There is no escaping taxes (unless you live in Monaco?). Tax assumptions are easier to build into Monte Carlo simulation because they are black and white, somewhat at least. We know the tax code right now, but might that change in the future? Case in point is the TCJA (Tax Cuts and Jobs Act) of 2017 that is expected to sunset in 2025. This would increase taxes on almost all Americans, but will it actually sunset? This remains unknown for the time being, but must be built into tax assumptions. Taxes are one of the biggest cash outflows for many people and inaccurate income and asset data can cause tax liabilities to be wildly off the mark.

Magnitude of Failures: Would you rather run out of money when you’re 70 or 100? Monte Carlo simulation treats both of these “failures” equally, but in reality the failures aren’t the same. I would consider running out of money at age 70 a catastrophic failure, whereas running out of 90, 95 or 100 is manageable. The latter situations typically only require small, temporary, changes to stay on track. Most of the time these adjustments are made early in retirement years when the timing of retirement coincides with poor market conditions. This is called “sequence of returns” risk: Portfolio values are suppressed because of market conditions, and portfolio withdrawals to support retirement spending simultaneously start. Temporary tweaks to retirement date and spending can usually negate sequence of returns risk regardless if Monte Carlo deems it a failure.

The “Probability of Success”

I have a love hate with the Probability of Success metric. On one hand it provides a statistical, data driven, figure surrounding the health of our financial plan. On the other hand, what does the number mean? And more importantly, what is a good score?

I like to reframe the Probability of Success; I also refer to it as “the likelihood retirement spending would need to be adjusted.” Let’s take a 70% success rate as the baseline for this example. In the Monte Carlo, 30% of the simulations failed and the retiree ran out of money. But remember, only small tweaks are needed in some of these failures to keep the plan on track. I would also refer to a 70% probability of success as a 30% chance spending would need to be adjusted at some point, whether temporarily or permanently.

The level of certainty that you will never have to adjust spending habits is a highly personal decision. To this extent, “what is a good score?” is an ambiguous question. People that value level income with certainty and keeping their legacy intact for heirs, should aim for a higher level of success, upwards of 90% or more. People that value higher income early in retirement and in turn may have to reduce income later in their plan, can aim for a lower level of success. The lower the success rate, the higher the chances of spending adjustments in retirement.

REd Flags

Beware of advisors using financial planning software as a sales tool. Nudging up return assumptions, reducing inflation, and leaving out fees, are just a few of the ways I’ve seen advisors pump up the probability of success rate to win a new client.. If you’re working with a financial advisor that is using financial planning software with Monte Carlo simulation, watch out for these things.

  1. If the financial advisor is also managing your assets, be cognizant of your actual portfolio return vs the assumed return being used in the Monte Carlo simulation. Financial planning software(s) does not use your historical performance or real portfolio holdings. The program uses returns of generic asset classes, such as Large Cap Growth, Small Cap Value, and International, just to name a few. And I’m not talking about small differences in returns; I saw an advisor at a firm showing a client a 7% return on their moderate portfolio in the Monte Carlo simulation, when in reality the advisor’s actual return since inception was just over 2%. Yikes.
  1. Fees. Financial advisors typically charge management fees that can be north of 1%. They may also be utilizing active mutual funds with high expense ratios upwards of 0.5-1%. Both of these fees should be baked into the return assumptions utilized in the simulation. Financial planning software does not automatically account for these fees in the assumed returns; it is up to the advisor to accurately apply their fees into the software. Make sure your advisor is showing you investment returns NET of fees.
  1. Your assumed return should reflect your actual portfolio. If you are a moderate investor with a 60/40 moderate portfolio of stocks and bonds, respectively, your assumed return should be reflective of that portfolio and not an aggressive allocation. Make sure the allocation your financial advisor is using reflects your ACTUAL portfolio allocation and risk tolerance.
  1. The probability of success is not a guaranteed outcome. It’s merely statistical modeling using a plethora of assumptions to make hypotheticals about the future. What could go wrong? Be suspicious of a financial advisor who makes the probability of success figures seem absolute. A lot can change over 10, 20 or 30 years. Make sure your financial advisor is clearly and honestly communicating how Monte Carlo simulation works.

Whether your goal is “Die with Zero” or leave a big inheritance for your kids, grandkids or favorite charity, Monte Carlo simulation when paired with real ongoing financial planning is the best avenue to achieve your desired outcome. Continuing to adjust, adapt and monitor that plan over time will help you to get the most out of your unique vision for retirement.