Earlier this year, the Wall Street Journal published an article titled “New Study Questions Retirement Planning Calculators’ Accuracy.” It reported on an academic study of 36 retirement planning tools. The research concluded that “in most cases, the available offerings are extremely misleading.”
This wasn’t news to me. My first article on the inaccuracies in retirement calculators dates back to August of 2012. And I’ve published an entire series on retirement planning software since then, including comprehensive reviews of free and paid retirement calculators.
Depending on where you are in retirement planning, the accuracy of the software you are using may not be a great concern, yet. Initially, your use of these tools will be more about your trajectory than your end point. In early career, you mostly need to know if your money moves are taking you in the right direction, and roughly when you might arrive at financial independence.
But, for near-retirees and post-retirees, accuracy in retirement modeling gets much more important….
For a major life decision like retirement, you can’t afford for your calculator to be off by hundreds of thousands of dollars. And, I’ve seen, first hand, over and over again, that this can happen easily with retirement calculator output. I routinely see 30-100% variations in calculator results. And even larger errors are not uncommon.
Those variations could represent years of extra, unnecessary work. Or, they might mean running out of money before you run out of life. Either kind of error is extremely serious for your future.
Even after making the retirement decision, accuracy is important. If you have substantial assets in retirement, then accurate, detailed tax calculations are essential if you intend to do any tax planning related to your tax bracket, RMDs, and potential Roth conversions.
What are the key factors working against retirement calculator accuracy? We can identify three sources of error: (1) design limitations and bugs in the calculators themselves, (2) inherent difficulties in predicting future variables such as investment returns, inflation rates, and life expectancy, and (3) user misunderstanding and mistakes.
I’ve covered difficulties in predicting future variables and user errors in other articles on the blog. In this article, I’d like to focus on how the calculators themselves can be designed and implemented for better accuracy.
My interest and knowledge in this area has grown dramatically thanks to Stuart Matthews of Pralana Consulting, one of the leading experts on retirement modeling. He first approached me in the fall of 2013 with his vision for higher-fidelity retirement calculations. (By fidelity we are referring to how well each calculator can potentially reproduce reality — the realism of its simulation.)
Many emails and discussions later, Matthews has largely convinced me that the best retirement calculators can and should attain higher accuracy, by avoiding obvious calculation errors. Yes, variables like investment returns and life expectancy are unknown. But there is no excuse for getting taxes wrong in a retirement simulation, for example.
Says Matthews: “Errors matter and they do compound just as interest does, and this can make a very substantial difference in long term projections. Low-fidelity calculators can lead to significant avoidable errors. It is possible to build high-fidelity tools that avoid the avoidable and restrict the uncertainty to the areas where it is truly unknown, such as inflation, market performance and life expectancy. And the very best tools acknowledge the uncertainty and present a range of possible outcomes based on what we do and what we do not know.”
Matthews has backed up his opinions with action, starting with his family of Pralana Retirement Calculators, and extending to his generous technical assistance to me and others working in the field. In fact, for many months now, he and I have been working closely on retirement calculation fidelity. You will begin to see the results in this article, and several to follow, that build on his extensive, hands-on experience with retirement modeling, coupled with my viewpoint as an early-retired software engineer.
We both agree that the future cannot be predicted accurately. But overly simplifying your attempts to model it can result in clearly misleading outputs and wrong conclusions. What follows are some of the most important calculator features we’ve identified to avoid errors and deliver the most accurate possible retirement planning results. If you’re a potential or current retiree, beware that your chosen financial planning tool includes all of these features, at least if you intend to make major life decisions based on its output!
Realistic Cash Flow
Let’s start with the essence of financial modeling: Money comes in as income. Money goes out as expenses. Some amounts change over time. Expenses and income generally increase, until retirement, but not always. Taxes take a bite. Low-fidelity retirement calculators oversimplify these details. High-fidelity calculators provide full resolution during both the accumulation (pre-retirement) and distribution (post-retirement) phases.
Consider your earned income. Do you enter gross or net into the calculator? A rough number is fine for rough results. But for accuracy, a high-fidelity retirement calculator will take a gross number from you and compute payroll taxes for Social Security and Medicare, whether you’re an employee or self-employed. The calculator will also handle income tax withholding and contributions to retirement accounts, plus any company matching.
Some low-fidelity calculators ignore separate expenses and income, simply assuming that what is left over, your retirement savings, is all that matters. But this is a crude approach, awkward and inaccurate for modeling real lives over time. And it’s wholly inadequate for the distribution phase, when it’s necessary to have an accurate assessment of your expenses in order to model retirement withdrawals. Calculators that want to assume your retirement income requirements are simply a percent of your working income are laughably inadequate in the real world, highly unlikely to deliver accurate answers for your individual situation.
In research published on the Pralana web site, Matthews shows the potential errors introduced into calculated results when users are asked to specify only the average savings contributions per year. Over time, that number isn’t likely to capture the actual difference between income and expenses. For example, Matthews reports that a difference of $5,000 over a 15-year accumulation period can lead to a 20% error in portfolio ending value after a 30-year retirement. Errors of this magnitude are to be expected when you consider trying to distill the buying and selling of cars, financing of college educations, and paying off a mortgage into a single annual “savings” number, as many low-fidelity calculators do.
Finally, when it comes to retirement income, an accurate retirement calculator must allow for precise input of the most common source for most people: Social Security. Your starting ages for Social Security can currently vary within a span of 8 years (ages 62-70) for each partner, with large implications for lifetime income. There are no general rules of thumb, and no substitutes for accurate calculations. Calculators need to handle any legitimate start age for Social Security and adjust, or allow you to adjust, the benefits accordingly.
Separate Account Types
In the modern world, most of us keep our savings in accounts at various financial institutions, not stuffed in mattresses, or buried in treasure chests. But not all accounts are created equal: For starters, there are three primary types based on their treatment at tax time: taxable accounts, tax-deferred accounts, and tax-free accounts.
As the name implies, the growth and income from taxable accounts are taxed, though often at different rates: Interest and dividend income are taxed at your income tax rate, but qualified dividends and long-term capital gains are usually taxed at lower rates. Once taxed, those amounts become part of the “basis” in your account, which a high-fidelity calculator will track.
The withdrawals from tax-deferred accounts are taxed at your prevailing rate for income. Tax-free accounts are exactly that: never taxed.
A high-fidelity calculator will compute tax liabilities for all three types of accounts precisely: There is no guesswork involved in these calculations, just a lot of number crunching. Lower-fidelity calculators usually can’t tax accounts differently, and rarely if ever track capital gains, though these distinctions could be important to your bottom line.
Matthews has studied differences in final savings balance as a function of how your initial balance is split between tax-deferred, tax-free, and taxable savings accounts. He found significant differences in final account balances depending on whether the calculator uses separate account types, or simply treats them as a common pool of money: He found up to an 18% variation, depending upon how the money is split between the various types of accounts at the start of retirement. He’s also looked at the long-term effects of modeling growth as simple interest versus capital gains: He saw up to a 19% variation, depending on whether the growth of regular savings takes one form or the other.
All three account types grow over time, but how? That’s dictated by the types of assets they hold: cash, bonds, stocks, real estate, and other more exotic asset classes. The exact mixture of those assets is defined by an asset allocation. High-fidelity calculators will allow you to define the types and quantities of assets held in each of your accounts, along with their returns in at least four categories — interest, dividend, realized gain, and unrealized gain — for accurate modeling of growth.
Over time, these assets growing at different rates will make up different proportions of your wealth. The solution often advised for reducing the ensuing risk is to perform regular rebalancing. There are several different styles of rebalancing available: periodic, or at a certain percent band. Does your retirement calculator handle your rebalancing style? If not, your results could be significantly distorted. In one study I did, introducing annual rebalancing produced about a 50% change in ending net worth!
To get money out of your accounts to cover retirement living expenses, you must make withdrawals. In the world of retirement modeling, this is much more complicated than inserting your debit card in an ATM.
In retirement, you aren’t simply consuming a paycheck that comes into a single account. Instead, as we’ve discussed, you are generally living off savings from a collection of accounts. So, you must regularly choose the source of your funds. And, as cash runs low, you must choose which investment(s) to liquidate.
High-fidelity retirement calculators let you specify the withdrawal order for your accounts: Which assets to tap first, and which ones next. And, because of the different growth rates and tax rates applied to your different accounts, your calculator can introduce major errors if it doesn’t match your real-world withdrawal pattern.
What if you inadvertently or intentionally withdraw from a retirement account before age 59-1/2? Without special care, you’ll incur substantial early withdrawal penalties of 10% of the amount withdrawn. A high-fidelity calculator will warn you and, if at all possible, help you design a different withdrawal sequence that doesn’t result in wasteful penalties.
For most traditional retirement accounts in the U.S., starting at age 70 the government requires you to make withdrawals and pay taxes, known as Required Minimum Distributions (RMDs). The exact amount is computed using a table with different ratios for each year, based on your average life expectancy. A high-fidelity calculator will get the timing and amounts right, automatically, so you don’t have to think about the details.
To understand the impact of RMDs on retirement wealth, I analyzed a 55-year long scenario starting with a couple age 40 and 35, making about $150K annually between them, retiring around 60, and collecting an average amount of Social Security. Disabling RMDs for this couple resulted in an ending net worth 67% higher than the actual, real-world value! Lesson learned: For detailed retirement cash flow and tax strategy planning, ignore RMDs only at your own peril. If your retirement calculator doesn’t handle them, get another calculator.
Whether the issue is withdrawal order, penalties, or RMDs, you can’t afford a low-fidelity calculator that ignores the details. “Accuracy is a function of the granularity with which the tool models real world accounts, including withdrawals from those accounts, especially in a predominantly withdrawal-type scenario as most people experience in their retirement years,” says Matthews.
Detailed Income Tax Calculations
We’ve saved the most difficult topic for last. It can be very tempting to ignore federal income tax calculations when it comes to retirement planning, because the complexity is overwhelming. That’s why the majority of lower-fidelity retirement calculators use a single, simplistic effective tax rate instead of performing detailed tax calculations with inflation-adjusted marginal tax brackets. The problem with effective tax rates is that they put the burden on you, the user, to distill your tax situation into a single, artificial number. For serious retirement planning that’s unrealistic and can be dangerously inaccurate.
In my recent post on effective tax rates I showed that the effective rate over the course of one typical retirement varied from about 24% to 0% to 8% to 9%. Which one was right? All of them, and none of them. Attempting to use an average effective rate produced a difference between the ending net worth determined by a high-fidelity calculator of $300K!
In his similar analysis, Matthews found that attempting to guess at effective tax rates can lead to magnified errors. In one of his studies, varying effective tax rates introduced 8% errors in the final savings balance for each 1% input error! Can you guess your effective tax rate for each year of retirement within 1%? Very unlikely. A simplistic numerical average is simply inadequate to accurately capture the complex U.S. tax code.
Even if you could guess at an accurate effective tax rate, Matthews points out that calculators vary in the ways they use that number, so even more caution is advised: For example, some tools allow you to specify the taxable percentage of each individual income stream, such as SS benefits which are either 0, 50 or 85% taxable depending upon your income. Also, does your tool apply the tax rate to your gross income or to your income after contributions to tax-deferred accounts have been eliminated? Another twist is that some lower-fidelity calculators ask you to specify net income — with taxes already removed. The chances of determining and using effective tax rates accurately on your own are slim.
Lastly, aside from tax rates, there are a bevy of special tax rules that can materially impact retirement wealth: Consider the various thresholds not indexed for inflation that can impact even middle-income retirees: the Medicare surtax, the Net Investment Income Tax, and the thresholds for taxation of Social Security income. Another couple of tax provisions potentially impacting retirees include the additional standard deduction at age 65, and the Personal Exemption Phaseout. None of these provisions can be adequately captured by using a simplistic effective tax rate. The best solution, if you need accurate retirement numbers, is to find and use a calculator that correctly performs detailed tax calculations.
The best retirement calculators are necessarily more complex. How do you know if they’re right? I’ve been programming computers for about 35 years now. Early on I learned that “any software that hasn’t been tested, doesn’t work.” As a rule, software cannot be trusted beyond the extent that it’s been verified. That means you, or some third party, needs access to the detailed results from any financial software worth trusting. You need to see all the inputs, and all the outputs, for each year of a simulation, with nothing left to assumption. This means all values used in the calculations, whether explicitly input by the user, or not.
In my opinion, the gold standard for verifying personal finance calculations is a transaction log. This is a report or output file showing, for each year in the simulation, all account balances at the beginning of the year, each transaction against those accounts, in chronological order, and the resulting account balances at the end of the year. While most retirement calculators do offer a tabular report, very few offer enough detail that you could theoretically hand-verify their calculations. And, even if you don’t intend to do that, the presence of such a file is testament that the developer is serious about accuracy. Without that data, a calculator really can’t claim it has traceable, verifiable results.
If you were audited by the IRS, you would be expected to produce proof of every single taxable or tax-deductible transaction for the accounts in question. Though retirement modeling doesn’t operate at that exact level of detail, the principle is the same, and the results are just as important. Typical low-fidelity retirement calculators make hidden numerical assumptions and show you only partial results. By contrast, the best high-fidelity calculators show you, in a readable report for every year of the simulation, each financial sum that is adding onto or subtracting from your bottom line. No detail is left to the imagination.
So we’ve reviewed five critical design features that qualify a retirement calculator or financial planning software as a serious modeling tool you can entrust with one of your life’s most important decisions. To get accurate retirement planning numbers, at a minimum, you need realistic cash flow, separate account types, accurate withdrawals, detailed tax calculations, and reporting of all intermediate results.
As Stuart Matthews and I have demonstrated with our individual research, if only one of these features is missing, your results can easily be wrong by tens of percent. If several are all missing together, as is the case in many low-fidelity retirement calculators, your results could be off by hundreds of percent. The answers will be essentially useless for detailed retirement planning. You’ll be flying blind as you make decisions involving potentially hundreds of thousands of dollars, and your happiness and security in later years.
Note: Stuart Matthews of Pralana Consulting and I are engaged in an informal technical collaboration aimed at raising general standards for accuracy in retirement modeling. We have no business relationship, no compensation arrangements, and no commercial intentions together.