Models of spending as retirement progresses

From Bogleheads
Jump to navigation Jump to search

This article examines in detail five models currently being used to represent retiree spending patterns from retirement until death. These models extrapolate the Initial Spending estimates developed for the beginning of retirement so that they describe spending until death.

The Constant (Real) Spending model assumes that the real spending at the start of retirement continues unchanged until death. Although this approach is widely used, it doesn’t correspond to the typical spending patterns exhibited by retirees. It usually (but not always) leads to an overestimate of the savings needed at retirement. The Stages of Retirement model breaks up retirement into 3 or 4 age ranges, each of which has a different real spending pattern. Spending within each range can either be constant or varying. This model does a much better job of representing the diverse spending patterns exhibited by retirees.

Investment Returns Dependent models allow spending in retirement to vary based on how total savings or investment returns change over time. These models are best combined with Dual Budget models of initial retirement spending. These models mimic the tendency of retirees to vary their spending based on their net worth.

Flexible Spending models allow numerous individual categories of spending to be separately incorporated into the total spending plan. Each spending category can be assigned its own start and stop age, as well as spending amount. This model is excellent at representing large step changes in spending during retirement. Many retirement calculators will combine elements of this model with others already described. This combination supplies a very flexible framework for accurately describing a diverse range of retiree spending patterns.

Life Cycle models calculate how a household should smooth their discretionary spending over their entire lifetime, not just in retirement. They provide a detailed spending and savings plan that should be followed in the pre-retirement years. Doing so will provide an optimized, real spending lifestyle for both the pre-retirement as well as the post-retirement years.

All of these model types have been implemented in actual retirement calculators. Specific examples of each can be found in the article Retirement calculators and spending.


This article examines in detail various models being used to describe how retiree spending changes from retirement until death. The material here is an in-depth expansion of the section Spending as Retirement Progresses from the introductory article on retirement spending models. Please also refer to that article for background information concerning approaches used to model Initial Retirement spending. These initial spending models supply the starting point for spending extrapolations across the retirement lifetime.

As in other modeling articles of this series, this article treats retirement spending as the key independent variable in the retirement planning process. The future retiree first develops an estimate of their desired spending in retirement. From there, the planning process makes assumptions about longevity, inflation and investment returns, and then estimates the future savings required to potentially sustain that retirement spending.

This article considers only the real component of retirement spending.[1] Inflation adjustments and how they are incorporated into the retirement planning process are covered in the article Inflation and retirement spending.

Retirement calculator models

There is a strong connection between retirement spending models and retirement planning calculators. Most software used for retirement planning, whether explicitly or implicitly, is written assuming some model of retirement spending. The adjustable options included in the software reflect the choice of spending model. It is an unfortunate fact that some retirement calculators have great strengths in many aspects of their design, yet are needlessly weak in how they incorporate retirement expenses.[2]

The article Retirement calculators and spending contains an extensive list of calculators which have been categorized in terms of the retirement spending model(s) they use. The tables in that article particularly highlight models of spending as retirement progresses. Refer to that article to find real-world examples of the spending models discussed here.

Retiree spending: key observations

The article Surveys of retirement spending reviewed and summarized data on retiree spending patterns that have been collected by two ongoing, national U.S. surveys: the Bureau of Labor Statistics' Consumer Expenditure Survey, and the Health and Retirement Study. From these surveys, as well as from the practical experiences reported by professional financial planners, certain key facts have become clear:

  • The average retiree exhibits a slight drop in real spending at retirement, followed by a steady decline in real spending as they age into their late 70’s.[3] [4] This decline in real spending is voluntary and not a result of limited financial resources.[5]
  • A substantial fraction of retirees (25%[6] to 28%[7] before the recent recession, rising to almost 4 out of 10 now[8]) enter retirement involuntarily. They exhibit a sharp drop in real spending at retirement.[6] If the involuntary retirement was health related, such retirees may subsequently exhibit a medical expense induced, real increase in total spending near their death.
  • A smaller percentage of retirees (roughly 12%[9]) exhibit an increase in real spending at retirement.[10] [11] This is often driven by a jump in travel or other leisure activities. After a certain time period these special activities end, and real spending drops to a lower level, often closer to that of “average retirees.”

The ability of each spending model to represent these key retiree spending patterns will be an important topic in the discussions.

Constant (real) spending models

This is the simplest of all approaches used to model spending as retirement progresses. The real spending at the time of retirement is assumed to continue unchanged until death. Inflation is not ignored, but as explained in the Introduction, it is treated as an separate, adjustable variable in the retirement planning process. From a nominal spending perspective, this model states that spending for any year in retirement is equal to spending in the first year times an independent inflation adjustment.

The constant spending model is most often combined with a Replacement Rate or Single Budget model of initial retirement spending. When a Dual Budget model of initial spending is used, it tends to be combined with a model that allows real spending to vary between the essential and preferred budgets as retirement progresses.

Comparison with retiree spending

Constant real spending models do not reflect the reality that the average retiree’s spending steadily drops as retirement progresses. Many retirement planners have pointed this out as a major shortcoming.[5] [12] They have also pointed out that such models lead to an overestimation of the total savings needed at retirement.

A constant real spending model could conceivably be appropriate in cases where retirement is involuntary and financial resources are limited. This situation would force the retiree to immediately drop down to an essentials only spending budget. To the extent that they really are living at their minimum acceptable level, their nominal spending would be expected to grow at somewhat the rate of inflation (constant real spending) as they were forced to pay ever increasing market prices for those essential goods. However, such a retiree would also be motivated to substitute, when possible, less expensive items for their essential budget spending needs. Such a Substitution Effect [13] would cause spending to somewhat lag behind constant real spending.

Relationship to withdrawal studies

Some of the earliest research on Safe Withdrawal Rates used constant real withdrawals in their calculations.[14] Constant real withdrawals is a close proxy for constant real spending that omits the complication arising from taxes. The constant withdrawals constraint allowed the researchers to focus on their main issue of interest: how Sequence of Returns Risk [15] impacts retiree savings survival. But because so much subsequent research has maintained the constant withdrawals constraint, some have drawn the mistaken conclusion that this is a realistic approximation to how retirees withdraw and spend money. But as previously emphasized, the average retiree exhibits real spending that steadily drops as they age.

Use in retirement calculators

Constant spending is the most commonly used model in free Internet retirement calculators. This is especially true for calculators that are deterministic, less so for those that utilize Monte Carlo or Historical Returns calculational approaches. It is used both for calculators that predict the target retirement savings starting from a desired retirement spending, as well as in calculators that proceed in the opposite direction.

There are free Internet retirement calculators that allow you to choose models other than constant real spending. But these will often include a constant spending model for purposes of comparison with simpler calculators.

Retirement calculators offered for sale are much less likely to rely solely on a constant spending model. But they often include this model for comparison purposes.

Although Life Cycle spending models sometimes yield a constant real spending in retirement, they do so from an entirely different, Economics Perspective.[16] This approach is discussed in its own section below.

Refer to the article Retirement calculators and spending for examples of retirement calculators that incorporate a constant real spending model.

  • Easy to understand, so it’s useful for illustrative (teaching) purposes.
  • Easy to implement in a retirement calculator or spreadsheet.
  • Does not match the reality of how average retirees spend money as they age.
  • Often leads to an overestimation of the total savings needed at retirement.

Stages of retirement models

Many studies have recognized that a typical retirement can be divided into 3 to 4 Stages or Phases.[17] Within each Stage retirees tend to exhibit similar patterns of physical activity and spending.[note 1] As a specific example, consider the following 4 Stage proposal by Robert Carlson: [18]

Budget Stage Description
Stage 1 First few years of retirement. Higher spending than in pre-retirement as the retiree pursues leisure time dreams while still having relative youth and good health.
Stage 2 Until about age 75. Spending settles back down to a lower, stable value.
Stage 3 After age 75. Spending shifts down again. The BLS Consumer Expenditure data suggest a 25% drop in real spending relative to early retirement.
Stage 4 End of life. Medical / long-term care expenses might drive up spending unless insurance is available to cover these costs.

Not all retirees pass through Carlson's Stage 1 of increased real spending early in retirement. The average retiree doesn't, but some retirees do. Typically retirees pass though Stages 2, 3 and 4. (These are the only stages considered in most 3 Stage models.) Some retirees in very poor health might experience increased real spending in Stage 4, but this depends on whether their insurance is able to cover the associated higher medical costs.

Many authors besides Carlson have chosen to use age 75 as one of their stage boundaries.[10] [12] [19] There are likely several reasons for this. For retirees having reasonable health, this age is approximately when natural aging processes lead to a slower pace of living (Michael Stein's Slow-Go stage [10] ). It is also possible that this age is chosen to match the BLS Consumer Expenditure tables, which break down their retiree data into age groupings of 65-74 and 75+.

Other authors feel that the diversity in retiree life experiences and associated spending prevent fixed ages from being assigned to the stage boundaries.[17] An example commonly cited is involuntary retirement. For those entering retirement involuntarily with insufficient financial resources, their spending could very quickly drop down to a low, Stage 3-like level even though they could be in their early 60's. Keeping such diversity in mind, a good Stages of Retirement model should allow the boundary ages to be adjustable.

Stages having constant real spending

In the very simplest Stages of Retirement model the real spending within each Stage is assumed to be constant. All changes in spending occur at the boundaries between stages. For the average retiree spending drops at these boundaries. This is especially true at the commonly used age 75 boundary. If a budgeting approach is used to estimate the real spending within each Stage, these can be referred to as Step Change budgets. But a changing (usually dropping) replacement rate could just as easily be used to model spending in each stage.[5]

More realism can be brought to this type of model by further breaking down spending into budget categories within each stage. In his Age Banding model [19] Somnath Basu incorporates a separate budget for each of 4 broad expense categories: taxes, basic living, healthcare and leisure.[20] At each of the stage boundaries his model allows a step change in real spending, either upwards, downwards or no change. For example, basic living expenses could be given step drops at each stage boundary, but healthcare given step increases at each boundary. Further, his model allows for a different inflation rate to be applied to each spending category within each stage.

Stages having variable real spending

Numerous studies have pointed out that retirees steadily reduce their spending as they age rather than exhibiting in a few sharp drops in spending. Based on this observation several retirement planners have suggested that Stages of Retirement models be made even more realistic by allowing real spending to vary gradually within one or more of the stages.

Tacchino & Saltzman. A very widely cited paper that suggested incorporating a steady spending drop into retirement planning calculations was published in 1999 by Kenn Tacchino and Cynthia Satlzman.[5] Arguing from the U.S. Consumer Expenditure survey data, they concluded:

"...all available data and research indicate that there is a gradual reduction in spending starting shortly after retirement. Spending decreases are natural, voluntary and acceptable, and should be reflected in the client's accumulation model. The downward adjustment in spending by age 75 is approximately 20 percent of the initial spending levels during retirement that starts at age 65."

This 20% reduction (2.2% annually) is an average across all retiree income levels. They also show that real spending drops for every income range within the BLS survey data, although the percentage decrease varies in a somewhat random fashion.

Tacchino and Saltzman didn't particularly emphasize using a continual (annual) real spending reduction until age 75. They spent more time presenting ways in which stage models having constant spending could be modified to approximate the steady spending decreases. But they did state that a retirement calculation incorporating a real spending drop each year would lead to a better estimate for the total savings needed at retirement.

Ty Bernicke's Reality Retirement Planning. Ty Bernicke published a widely cited paper Reality Retirement Planning: A New Paradigm for an Old Science in 2005.[12] In it he emphasized using a steady, real spending reduction in each successive year of retirement until age 75. Real spending was assumed to remain constant in stages after age 75. His arguments were (again) based on data from the BLS Consumer Expenditure Surveys.

In his paper Bernicke recommends using a 50% spending drop between the 55-59 age group and age 75. A sample calculation presented used an annual real spending drop of 3.7%. This is certainly too large. Referring back to Table 1 in this paper, it can be seen that the raw Total Consumer Expenditure dollar amounts from the 2002 CE Survey were used. These dollar amounts included savings dollars (Social Security payroll taxes and personal retirement savings) which should have been subtracted out before estimating the percentage spending drop.[21] Also, there was no correction for the different average household sizes between the age groups (i.e. 2.1 persons for age 55-59 versus 1.5 persons for age 75+). Incorporating these corrections would have reduced the predicted spending drop by age 75.

Recommended Annual Spending Drop. Bernicke's annual percentage drop of 3.7% until age 75 is too large. The equivalent annual spending drop from Tacchino and Saltzman, 2.2%, is much lower and probably closer to the truth. Fisher have analyzed the same Consumer Expenditure Survey data and recommended using a 1% annual real spending drop until the late 70's.[3]

The BLS Consumer Expenditure Survey doesn't supply the best database for analyzing drops in real spending as retirement progresses. This is because the BLS survey doesn't follow the same households in each successive year. Changes in spending patterns with aging can only be derived by comparing shifting averages in large populations of households. It would be much better to perform retiree spending studies using the Health and Retirement Study survey database. In this biennial survey the same households are followed from age 51-56 until death. Unfortunately no such study of retiree spending patterns appears to have been performed.

William Bengen’s Prosperous Retirement. Published in 2001, Bengen’s Prosperous Retirement model [22] is aimed at those retiree’s whose spending early in retirement is higher than average. This is in contrast to the previous two models that targeted a more "average" retiree. This model has 3 Phases: Active, Transition and Passive. In the Active Phase (until age 75), the model assumes a constant real spending. In the Transition Phase the model has a strongly declining real spending each year until age 85. (The example in the paper used 4% below inflation annually.) In the Passive Phase the model has a weakly declining real spending each year until death. (The example in the paper used 2% below inflation annually, which gave a slightly increasing nominal spending each year.)

Comparison with retiree spending

The use of 3 (or 4) Stages of Retirement, each of which can have a different real spending, allows for much better modeling of retiree spending patterns. All three of the basic patterns discussed earlier in the Key Observations section can be mimicked. And if the first one or two stages allow a gradual annual spending drop rather than just a constant real spending, then an even better match to actual retiree spending can be achieved.

Use in retirement calculators

Calculators having a Stages of Retirement spending model are much less common than those using a constant real spending model. Referring to the Calculators and Spending article, only a few of the calculators explicitly incorporate this model. Some of these have included Ty Bernicke's Reality Retirement Planning approach.

Although the documentation for only one of calculators explains how (the Flexible Retirement Planner), it is usually possible to set up a Stages of Retirement model with any calculator having a Flexible Spending capability. Each spending Stage (start age, stop age and constant real spending amount) is modeled using one of the custom spending inputs. To make this calculate properly, it may be necessary to set the initial spending budget to $0.

  • Allows the diversity in retiree spending patterns with aging to be incorporated into the retirement calculation.
  • Significantly improves model accuracy without excessively increasing complexity.
  • Additional work needed to estimate budgets or replacement rates for each Stage.
  • There is some uncertainty in the most realistic annual percentage spending drop to age 75.
  • Somewhat more difficult to implement in a retirement calculator than constant real spending.

Investment returns-dependent models

It isn’t surprising that retirees would cut back on their spending during times when they don’t feel as financially secure.[23] For retirees having savings invested in stocks or stock mutual funds, bear stock markets would be such a time of reduced spending. This is a natural response to declines in personal net worth caused by declining equity values. Conversely, when a retiree feels financially secure (e.g. during a bull stock market), they would naturally increase their real spending.[24] These periods of increased and decreased spending can be modeled by using Withdrawal Methods that contain a dependency on either the previous year’s total savings or its investment return. An extensive summary of these methods is given on the Variable Withdrawals in Retirement archived page from the old Bob’s Financial Website.

A Dual Budget model can be added as a natural complement to such variable withdrawal methods. The Essential Budget spending level would set a lower limit on spending as retirement progressed. The Preferred Budget spending level would set a corresponding upper limit on spending. These constraints would be included with those supplied by the withdrawal model when calculating the approximate savings needed at retirement. But since the Dual budget is an initial spending model, some method of extrapolating the spending across the retiree’s lifetime must be used.

It seems reasonable to extrapolate the Essential budget component of a Dual budget model using constant real spending. To the extent that a retiree really was living at their minimum acceptable level, their nominal spending would be expected to grow at somewhat the rate of inflation (constant real spending) as they were forced to pay ever increasing market prices for those essential goods.

Extrapolating the Preferred budget component using a constant spending model is not reasonable, for all the reasons previously discussed. Instead, the Preferred budget should be extrapolated using either a Stages of Retirement model or a Flexible Spending model. Both allow a more realistic description of how actual retirees alter their real spending as they age.

Comparison with retiree spending

By itself an Investment Returns Dependent spending model is not able to adequately describe the key observations on retiree spending discussed in the Introduction. It has no ability to match the steady decline in real spending as retirees age. But it does supply an element of reality that other models lack: the ability to reflect how retirees alter their spending in response to changes in their net worth. A combination of these two spending tendencies would be very beneficial.

A useful analogy is to think of waves on the ocean as the tide goes out. The waves would represent the investment return-dependent spending and the tide going out would represent the steady decline in spending as the retiree ages. The presence of the waves raising and lowering the ocean surface level will not nullify the more powerful effect of the lowering tide. But without including both effects, the water level at any instant can’t be fully described. Similarly, retiree spending in any year is better described by combining these spending models.

Use in retirement calculators

Calculators having an Investment Returns Dependent spending model are much less common than those using a constant real spending model. However, referring to the Calculators and Spending article shows that more calculators incorporate some variation of this model than those explicitly incorporating a Retirement Stages model.

  • Reflects the tendency of retirees to adjust spending up or down as their net worth changes.
  • Nicely complements variable withdrawal methods used to ensure savings survival in retirement.
  • More realistic than a Constant Spending model.
  • Doesn’t match the reality that average retirees spend progressively less money as they age.
  • Less able to mimic the broad range of retiree spending patterns than a Stages of Retirement or a Flexible Spending model.

Flexible spending models

A Flexible Spending model allows numerous individual categories of spending to be separately incorporated into the total spending plan. Usually there is no need to enter every conceivable budget category as a separate spending item. Rather, these models usually allow you to first enter a base spending amount via another model (Constant Spending, Stages of Retirement, etc.). Then any spending that still isn’t properly represented can be entered as a separate, individual spending item. Typically the starting age, ending age and spending amount for each category are the only required information.

This type of modeling flexibility is particularly important for retirees anticipating one or more large expenses that occur only over a limited age range. Thus it offers the potential for detailed personalization. Consider the following situations where a Flexible Spending model would allow a closer match between actual and modeled spending:

  • An early retiree plans to purchase expensive, individual medical insurance until they become eligible for a less expensive Medicare + Medigap combination at age 65.
  • A retiree wants to explore how possibly needing to pay for several years of her mother’s long-term medical care might affect how long her own savings would last.
  • A near-retiree with mortgage payments lasting another 13 years wants to incorporate this into his proposed retirement spending plan.
  • A person 10 years from retirement wants to accumulate enough savings to enable a winter home to be purchased in Arizona several years after retiring at age 65. The current home would be retained until age 75 and then sold.

A Flexible Spending model is able to incorporate such category-specific, step changes in real spending. At the same time other spending categories, for example basic living expenses, could be described using other models that provide a better match to fairly smooth spending changes.

Often a Flexible Spending model can be used to mimic the spending pattern of a Stages of Retirement spending model. This is accomplished by using one flexible spending category to contain all the spending within one of the retirement stages. Most Flexible Spending models assume constant real spending within each category, so the resulting Stages model would also have constant spending. But there are a few calculators that allow variable real spending within a flexible spending category.

If a retirement calculator provided a very large number of separate flexible spending entries, then one could even create a separate spending entry for each year of retirement.[25] This would provide enough planning flexibility to cover almost any conceivable situation.

Robert Carlson has discussed the advantages to being able to accurately model large, periodic expenses in retirement.[18] Examples of such periodic expenses are car purchases and major home appliance replacements. The simple approach is to use a “sinking fund”, where the total cost of the item and its expected lifetime are used to calculate an effective annual cost. But a more precise treatment would use a specially designed flexible spending category. This special category would accept the following data: the expense year, total cost, and expected lifetime. The spending model would then calculate all subsequent years in which a repeated purchase of that item would be anticipated, as well as the inflated cost in each of those years. Such an approach allows a more realistic projection of variable retirement spending. This level of modeling detail would enable tax bracket management to be incorporated into retirement planning.[26]

Comparison with retiree spending

Provided that there are enough flexible spending categories/entries available, this model can do an excellent job of representing the diversity of retiree spending patterns discussed previously in the key observations section. If there are only a limited number of categories/entries available (e.g. 3-5), then representing a steady decline in retiree real spending through their late 70’s or early 80’s would not be possible. At best a Stages of Retirement model equivalent with constant real spending could be created. But even in such a limiting situation the final spending estimate would be superior to that from a constant real spending model.

Use in retirement calculators

Calculators having a Flexible Spending model are much less common than those using a constant real spending model. However, referring to the Calculators and Spending article shows that a handful of free calculators do incorporate some variation of this model. This spending model is found in most purchased retirement calculators.

  • Capable of providing a personalized description of retirement spending for a broad range of retirees.
  • Best model for handling large changes in spending that might occur at irregular times during retirement.
  • Entering the personalized spending data can be very time consuming for some calculators.
  • Not well suited for mimicking investment returns dependent retiree spending.
  • Additional effort required to incorporate this model into a retirement calculator.

Life Cycle models

Life Cycle models calculate how a household should smooth their discretionary spending over their entire lifetime, not just in retirement.[27] Because they typically make a distinction between essential spending and discretionary spending, they are conceptually connected to Dual Budget models of retirement spending. Life Cycle models provide a detailed spending and savings plan that should be followed in the pre-retirement years. Doing so will provide an optimized, real spending lifestyle for both the pre-retirement as well as the post-retirement years.

Spending predictions from Life Cycle models may sometimes be similar to spending in Constant Real Spending models, because in a few cases both will recommend a constant real spending in retirement. But further investigation reveals a deep difference in their fundamental approach. A Life Cycle model predicts how a household should smooth their discretionary spending over their entire lifetime, not just in retirement.[16] Economic Utility theory predicts that people would be happiest if they neither over-spend nor under-spend each year as they age, whether retired or not. Starving now to ensure a comfortable retirement makes no more sense than splurging now and starving in retirement!

It is not a simple problem to estimate how much to save each year before retirement so that a comparable lifestyle could be maintained each year both before and after retirement. This is especially the case if uncertainties in future investment returns, life spans, and health care expenses are to be taken into account.[28] Calculators using the Life Cycle approach make use of a computational technique called Dynamic Programming to overcome these difficulties.

The following table summarizes the main differences between the Life Cycle model and other spending models: [16]

Life Cycle spending model Other Spending Models
Smoothes spending both before and after retirement Only focus on spending after retirement.
Calculates annual savings target for every year before retirement. These targets will vary each year based on other household expenses (mortgages, college costs, etc.). Suggest a target savings at retirement. Sometimes also suggest a fixed or inflation-adjusted amount to save annually. No coordination with other life expenses.
Capable of incorporating life insurance recommendations. Life insurance is not considered.
When to start Social Security can be optimally calculated. When to start Social Security is an adjustable, independent variable.
Can fully incorporate federal and state taxes into the calculation of optimum saving and spending schedules. Spending targets are set without regard for tax implications.

A Life Cycle model sometimes predicts a constant real spending during retirement, for example when longevity risk aversion is ignored and a fixed (deterministic) retirement lifespan is assumed.[27] But even in this restrictive case that spending level should not be regarded as an unchanging quantity to be spent each year regardless of how retirement investments perform. Rather, the Life Cycle model calculation should be run every year, so that the predicted optimal spending can be adjusted for investment performance.[16] In this sense a Life Cycle model is different from a constant spending/withdrawal model set up using a Safe Withdrawal Rates approach.

When a Life Cycle model is run using more realistic assumptions (e.g. a probabilistic retiree lifespan combined with longevity risk aversion), it predicts that the retiree should reduce their real spending as they age.[29] The lower the retiree's longevity risk aversion (i.e. fear of outliving their savings), the more they will pull real spending forward into the earlier years of retirement. Of course if such a retiree lived an unexpectedly long time, their real spending late in retirement would need to be greatly reduced. But stating that the retiree's longevity risk aversion is low means that they are little bothered by such a possibility!

Life Cycle models of retirement spending can readily incorporate pension-like sources of retirement income that can't be out-lived, such as U.S. Social Security benefits, Defined Benefit pensions, and immediate fixed annuity payments (SPIA's). When such sources of income are included, it is observed that the optimal Portfolio Withdrawal Rate (PWR) from the retiree's personal savings depends on their longevity risk avesion and on their level of pension-like income. The larger the amount of the preexisting pension income, the preater the PWR is and thus the greater is the optimal consumption rate.[29]

Comparison with retiree spending

If implemented using a fixed retiree lifespan and no longevity risk aversion, Life Cycle models predict a constant (real) discretionary spending target throughout retirement. This is clearly not in agreement with the spending patterns of the average retiree discussed previously in the key observations section. But when Life Cycle models are implemented using more realistic assumptions (e.g. longevity risk aversion and probabilistic lifespans), they correctly predict that real spending decreases with age.

Economic theory predicts that people should attempt to maintain a constant consumption rather than a constant spending.[6] For example, spending on food tends to drop sharply after retirement. However, careful studies have shown no change in food consumption, i.e. the amounts and quality of the food consumed remain the same. Because retirees choose to spend more time preparing their food, they therefore need to spend less money. Similar trade-offs doubtlessly affect other budget categories, especially as retirees make use of the ubiquitous “Senior Discount.” Ideally, a Life Cycle model should recognize this difference between smoothed lifetime consumption (the actual goal) and smoothed lifetime spending (a simple proxy).

As for retirees who plan on sharply increasing real spending after retirement for extensive travel and other expensive leisure activities, the Life Cycle model simply says: Why wait? Enjoy some of these activities while you’re young! The Life Cycle model will help you determine how much of the leisure activity expense can be pulled ahead before retirement.

Use in retirement calculators

Calculators incorporate Life Cycle models less frequently than other spending models. This is not surprising, given the complex programming required to fully implement the model. Some retirement calculators choose to partially implement the Life Cycle concepts, thereby simplifying the requried programming. The Calculators and Spending article shows only a few calculators that either fully or partially incorporate the Life Cycle approach. ESPlanner and ESPlannerPLUS fully incorporate the characteristics described above. Optimal Retirement Planner, being a free program, not surprisingly has only a partial implementation of these characteristics. But it is still a very powerful calculator.

  • Gives specific, annual spending and saving goals both before and after retirement.
  • Able to recommend optimum age to start Social Security.
  • Able to fully integrate federal and state taxes into the calculation of optimal spending.
  • Simple implementations that yield a constant real spending will not match average retiree spending patterns.
  • Complicated to fully implement into a retirement calculator.


  1. For example Michael Stein in his book The Prosperous Retirement: Guide to the New Reality (Emstco, LLC, 1998) defines his stages as Active Retirement Phase ("Go-Go"), Passive Retirement Phase ("Slow-Go") and Final Retirement Phase ("No-Go"). Although his stages primarily refer to physical capabilities, Stein also contends that changes in real spending also accompany the stage changes.

See also


  1. In economics real refers to the purchasing power net of any price changes over time. Refer to the Wikipedia article Real versus nominal value (economics)
  2. John Turner and Hazel Witte, Retirement Planning Software and Post-Retirement Risks, The Society of Actuaries (2009), 118pp. Exhaustively examines retirement software strengths and weaknesses, including weaknesses in the area of expenses.
  3. 3.0 3.1 J. Fisher, D. Johnson, J. Marchand, T. Smeeding and B. Torrey, The Retirement Consumption Conundrum: Evidence from a Consumption Survey, Center for Retirement Research at Boston College Working Paper WP 2005-14, (Dec. 2005), 30 pp.
  4. Yung-Ping Chen, John C. Scott and Jie Chen, Retirement Spending and Changing Needs during Retirement: Summary of Regression Analysis, presented at the 2007 Society of Actuaries Annual Meeting.
  5. 5.0 5.1 5.2 5.3 Kenn B. Tacchino and Cynthia Saltzman, “Do accumulation models overstate what’s needed to retire?”, Journal of Financial Planning, Vol. 12 (Feb. 1999), pp 62-74.
  6. 6.0 6.1 6.2 Erik Hurst, Understanding Consumption in Retirement: Recent Developments, Chapt. 3 in Recalibrating Retirement Spending and Savings (eds. John Ameriks and Olivia Mitchell), Oxford University Press (2008) pp 29-45.
  7. Marie-Eve Lachance and Jason Seligman, Involuntary Retirement: Prevalence, Causes and Impacts, Paper presented at the annual meeting of the Association for Public Policy Analysis and Management-APPAM (2009),47p.
  8. Anna Rappaport, “The Golden Glitch – Expanding Longevity and Shrinking Work Lives,” Contingencies – the magazine of the American Academy of Actuaries (Nov/Dec 2013) pp 36-41.
  9. Michael Hurd and Susann Rohwedder, The Retirement Consumption Puzzle: Anticipated and Actual Declines in Retirement Spending, NBER Working Paper 9586 (March 2003).
  10. 10.0 10.1 10.2 Michael Stein, The Prosperous Retirement: Guide to the New Reality, (Emstco, LLC, 1998), 312 pp.
  11. Robert Carlson, How to Vary Spending During Retirement, at Bob Carlson’s Retirement Watch.
  12. 12.0 12.1 12.2 Ty Bernicke, Reality Retirement Planning: A New Paradigm for an Old Science, Journal of Financial Planning, (June 2005), 8 pp.
  13. Refer to this brief description of the Substitution Effect at Investopedia.
  14. Examples are The “Trinity Study” (Philip L. Cooley, Carl M. Hubbard and Daniel T. Walz, “Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable”, AAII Journal, Vol. 20, Feb. 1998 ) and the early research by William Bengen (William P. Bengen, “Determining Withdrawal Rates Using Historical Data”, Journal of Financial Planning, Jan. 1994, pp14-24.)
  15. For a discussion about the origin and important of Sequence of Returns Risk, see Understanding Sequence of Returns Risk,, (archived).
  16. 16.0 16.1 16.2 16.3 Laurence Kotlikoff, Economics’ Approach to Financial Planning,Journal of Financial Planning (March 2008).
  17. 17.0 17.1 M. Cowell, R. Helman, A. Rappaport, S. Siegel and J. Turner, The Phases of Retirement and Planning for the Unexpected, Society of Actuaries 2007 Risks and Process of Retirement Survey Report (Aug. 2008), 26pp.
  18. 18.0 18.1 Robert C. Carlson, “How Much Will You Need”, Chapter 3 in The New Rules of Retirement: Strategies for a Secure Future, (John Wiley & Sons, Inc., 2004) 288 pp.
  19. 19.0 19.1 Somnath Basu, Age Banding: A Model for Planning Retirement Needs, Financial Counseling and Planning, Vol. 16, No. 1 (2005) pp29-36.
  20. Most budget models don’t consider taxes to be an independent budget item. Rather, they calculate taxes based on the income required to cover spending in the remaining categories.
  21. The BLS Consumer Expenditure Survey tables consider retirement savings to be an “expenditure”, just like spending on food and housing are expenditures. This makes sense from the survey’s perspective, since the goal is to account for every after-tax dollar received by a household. But from a retirement planning perspective, only the categories of spending that will persist throughout retirement are of interest.
  22. William P. Bengen, Conserving Client Portfolios During Retirement, Part IV, Journal of Financial Planning, (May 2001).
  23. Kevin J. Lansing, Gauging the Impact of the Great Recession, Federal Reserve Bank of San Francisco FRBSF Economic Letter, July 11, 2011.
  24. Lonnie K Stevans, Aggregate consumption spending, the stock market and asymmetric error correction, Quantitative Finance, Vol. 4 (2004) pp 191-198.
  25. Several of the purchased programs tabulated in the article Retirement calculators and spending can include unique spending or income events for each year of retirement.
  26. Such large purchases could require large withdrawals from a tax-deferred account (IRA, 401(k), or similar) to pay for them. In some cases these withdrawals would be sufficiently large to push the retiree into a higher tax bracket. A better option then would be to make a combination of withdrawals from both tax-deferred and Roth IRA accounts, such that taxable income was maintained within the lower tax bracket.
  27. 27.0 27.1 N.Charupat, H. Huang and M. Milevsky, Strategic Financial Planning Over the Lifecycle, (Cambridge University Press, 2012).
  28. John Scholz, Ananth Seshadri and Surachai Khitatrakun, Are Americans Saving “Optimally” for Retirement?, Journal of Political Economy, Vol. 114, No. 4 (2006), pp 607-643.
  29. 29.0 29.1 Moshe Milevsky and Huaxiong Huang, Spending Retirement on Planet Vulcan: The impact of longevity risk aversion on optimal withdrawal rates, Financial Analysts Journal, Vol. 67, No. 2 (2011) 14 pp.

External links