Many retirees have to decide whether to take their work pension as an annuity or a lump sum. While the value of the lump sum and annuity are straightforward to understand, comparing the two requires you to make assumptions about how many uncertain variables such as inflation, asset class returns, and longevity will evolve in the future. Any insights from this exercise then need to be combined with your unique financial situation and preferences to make the decision that is appropriate for you. The main purpose of this discussion is to provide an overview of the considerations for choosing between a lump sum and an annuity as well as illustrate how we approach answering this question. We emphasize the importance of a multifaceted approach to answering this question to reduce model-specific risk.
The decision of whether to choose a pension as an annuity or lump sum requires consideration of the following:
o the value of the annuity;
o any survivor benefits for the annuity;
o any cost of living adjustment for the annuity;
o the value of the lump sum;
o your risk tolerance;
o the age at which annuity payments begin;
o how far into the future you will begin to take the annuity or lump sum;
o your longevity;
o future inflation projections;
o the asset allocation of the lump sum portfolio;
o future return projections for the components of the lump sum portfolio;
o behavioral investment management risk;
o whether or not you would like to leave any amount to your heirs;
o whether or not you will receive an annuity supplement and the amount of the annuity supplement;
o how much of the annuity you expect to spend;
o the growth rate in the amount of the annuity you expect to spend;
o how much you would need to draw from the lump sum portfolio;
o the growth rate in lump sum portfolio withdrawals;
o tax considerations; and
o your overall financial planning situation.
Many of the considerations for determining whether to take a pension as an annuity or lump sum are fixed when making this decision for a given year. For example, the amount of the lump sum, the amount of the annuity, survivorship details, any annuity supplement, and the age at which annuity payments begin are such fixed considerations not subject to uncertainty. However, many other considerations have an uncertain nature. For example, it is not possible to know what the evolution of investment returns for various asset classes or inflation in the future will look like. Longevity creates another source of uncertainty.
It is these uncertainties that introduce difficulty in determining whether to take a pension as an annuity or lump sum. There are multiple approaches you can take to address these uncertainties. One possibility is to simply make an educated guess of the value of each of the uncertain inputs. This educated guess can be informed by historical averages for asset classes as well as life expectancy for an individual given that individual’s characteristics. However, this approach risks not giving a complete picture of the range of possible outcomes.
Another possibility is to supplement these best guesses with a variety of different scenarios. This certainly improves the analysis but is limited by your ability to accurately capture the range of possible outcomes in the future.
Also, another complication arises when you consider sequence of return risk. A given average annualized return for an asset class over a period of 30 years can be achieved by a large variety of sequences of 30 individual returns per year. For someone taking a stream of withdrawals from a portfolio over 30 years, large equity market declines at the beginning of the 30 year period will have a different impact than large equity market declines at the end of the 30 year period. For a static asset allocation between equity, fixed income, and cash, the existence of the withdrawals is what causes this difference because sequence of returns would be a nonissue for a given annualized rate of return on a portfolio in which no withdrawals were taken. Things are complicated further when you factor in an asset allocation that changes over time, such as by decreasing exposure to equity and increasing exposure to fixed income as you get older. In a case like this, even if there were no withdrawals from a portfolio, the sequence of returns for each asset class would matter because the allocation to each asset class changes over time.
Acknowledging the complications that arise from sequence of return risk for a portfolio experiencing cash flows that may be amplified by a change in asset allocation over time, the next question becomes how do we go about addressing this risk and others? Though past performance of an asset class does not predict future results, it can be helpful to use the historical experience of various asset classes as a starting point to help form expectations about the range of outcomes in the future. In fact, Monte Carlo simulations used by financial professionals typically rely on historical returns of various asset classes, the standard deviation of those returns, and correlations of those asset classes with each other to simulate the range of outcomes possible in the future. Unfortunately, such models can still fail to completely capture even the historical behavior of asset classes because of their reliance on just the average returns of asset classes, the standard deviation of returns for those asset classes, and the correlation of asset classes with each other. Those inputs are typically paired with what is called a normal probability distribution to simulate future returns. The problem is that the normal probability distribution is not appropriate for modeling financial assets in our opinion. The probability of extreme outcomes and the skew of returns for financial assets can deviate significantly from what would be predicted by a normal distribution. A better way to implement a Monte Carlo simulation would be to assume a parametric distribution that more appropriately represents the excess kurtosis (high likelihood of extreme outcomes) and skew (asymmetry) of financial asset distributions or that uses a nonparametric technique. Nonparametric techniques include the resampling bootstrap methodology pioneered by Brad Efron in the 1970s.
In light of the limitations of simple estimates and Monte Carlo simulations using an inappropriate probability distribution assumption, we have found that one of the most straightforward ways to capture the range of historical outcomes is simply to run a historical simulation that assumes starting in a given year and experiencing the sequence of asset class returns and inflation rates for a time horizon (e.g., 30 years) beginning in that year. Then, we continue this process for all subsets of the range of historical data. This is also called a backtest. We run this analysis on data going back to before the start of the Great Depression. Though the future can look very different than the past, such an approach can capture the range of outcomes experienced historically.
This range of outcomes includes the Great Depression from the 1920s through the 1930s; price deflation during the Great Depression; the expansion of the federal government during the New Deal in the 1930s; World War 2; a multi-decade Cold War with the Soviet Union; nuclear proliferation and the 1962 Cuban Missile Crisis; President Lyndon B. Johnson’s War on Poverty; the United States abandoning the gold standard and adopting fiat money; extremely high inflation combined with recessions in the 1970s and 1980s; Paul Volcker’s interest rate hikes in the 1980s; the dissolution of the Soviet Union; the late 1990s Asian Financial Crisis; the Dotcom bubble and bust of the 1990s and early 2000s; 9/11 and wars in Afghanistan and Iraq; the Global Financial Crisis of 2007 to 2009; the European Debt Crisis of the 2010s; Covid-19; post-pandemic inflation; and an unprecedented adverse market for “safe” long-term government bonds in 2022. Such an approach inherently captures scenarios few professionals today experienced in their lives and that, perhaps, may not be imaginable to the extent that it would show up in a scenario analysis. Also, such an approach fully captures the messiness of the historical experience in a way that we find superior to making a clean and simple normal probability distribution assumption.
The historical simulation approach can be run across different assumptions for longevity (e.g., dying at age 100, dying at age 90, etc.). This allows you to pair the historical experience of asset class returns and inflation with different assumptions for longevity.
In addition to running a historical simulation across longevity, it is appropriate to run a scenario analysis that involves varying many of the assumptions for comparing an annuity and lump sum to mitigate the risk of limiting your analysis to what has happened in the past. This can involve asset class returns that differ from historical experience, inflation rates that differ from historical experience, various sequences of returns, and the influence of adverse behavior on an investment portfolio. An example of adverse behavior would be failing to stick to a long-term investment strategy and can be simulated by a scenario such as a historical experience of starting right before the Great Depression in which you sold everything following significant market declines.
This historical simulation and scenario analysis can also be supplemented with a Monte Carlo simulation. Again, it is preferable that this Monte Carlo simulation addresses the limitations of using a normal distribution assumption.
Combining a historical simulation with a customized scenario analysis gives you a more comprehensive view of the range of outcomes relevant for comparing a lump sum and annuity in isolation. The next step is to combine this information with your overall financial planning situation and individual circumstances. This is where a detailed financial plan is crucial. For example, the attractiveness of an annuity versus a lump sum to a single person that has a significant amount of portfolio assets appropriately invested, no desire to pass on remaining assets to heirs, and a desire to reduce overall market risk can be significantly different than a married person that wants to pass on remaining assets to his or her children and is mostly concerned with the decision with the maximum expected inflation-adjusted lifetime value. Though the expected lifetime value of the lump sum portfolio may be more attractive than the annuity in a given case, the first person may find the annuity more attractive for its ability to outperform the lump sum over periods of adverse market returns and supplement that person’s existing asset allocation. Analyzing the difference in expected value between the two approaches allows such a person to make a decision in the context of how much expected lifetime value it would cost. Importantly, this discussion does not factor in any credit risk that may be relevant for the annuity. As always when using historical data, it is important to emphasize that past performance of an asset class does not provide any certainty regarding future results, there are no guarantees with investing, and it is not possible to predict the future with certainty.
Another area where a holistic approach involving a detailed understanding of your financial situation is important when considering whether to take a pension as a lump sum or an annuity is tax planning. For a lump sum that can be rolled over into an IRA and invested appropriately for the long run, there can be benefits from tax deferred growth of the principal, though withdrawals from the IRA will be taxed at ordinary income rates. This is in comparison to an annuity, from the pension of a company’s qualified retirement plan, taxed as received at ordinary income rates with no flexibility to adjust the stream of income based on your unique financial circumstances or invest an underlying principal amount. One thing to note is that a portion representing your contributions may be nontaxable when received as a lump sum or annuity. Planning opportunities such as Roth conversions provide additional considerations when it is possible to take a pension as a lump sum that can be rolled over into an IRA.
These holistic financial planning considerations illustrate an important aspect of our financial planning process: managing model-specific risk. No financial model, however sophisticated, can accurately represent the complexity of the real world and the unique goals, preferences, and circumstances of all economic participants in it. In fact, model risk management failures have been, in our opinion, a key ingredient of many historical financial calamities. A couple recent examples are the failure to appropriately model default correlations in mortgage derivatives during the Global Financial Crisis of 2007 to 2009 as well as the failure of some regional banks to manage interest rate risk in their holdings of “safe” government bonds in 2022 to 2023. Every model of the real world is at the mercy of the assumptions that went into it. In our view, there is an overreliance by participants in the financial industry on models developed externally by a third party in which the assumptions underlying such models are rarely understood, questioned, or modified. This is why we believe it is appropriate to supplement models focused on a specific problem, such as a pension analysis, with a detailed financial plan. By looking at the problem from multiple angles, you are more likely to mitigate the blind spots inherent in any individual approach.
The information in this blog post is for informational and educational purposes only. It should not be considered investment, tax, or legal advice. Please consult your investment, tax, and/or legal professional for such advice specific to your unique situation before making investment, tax, and/or legal decisions. Links to third parties have been provided as a courtesy to the reader to cite potentially helpful references. We do not claim that the material included in third party links is complete or accurate. Investing involves risk, including the potential loss of principal. There are no guarantees with investing. Past performance of an investment does not guarantee future results.
