Replacement rate models of retirement spending

Replacement rates (or ratios) provide a simplified method to estimate your spending needs in retirement. The gross (pre-tax) income you’ll need in retirement is modeled as equaling:


 * Gross Income (retired)  =  Gross Income (pre-retirement)  &times;  Replacement Rate

As is true for any simplified method, the predicted retirement income might or might not be correct for you. This depends on how closely your individual situation matches the assumptions used to derive the replacement rate used in the above equation.

The primary source of tabulated replacement rates for use by U.S. retirees is the GSU/Aon RETIRE Project Report. Aon Consulting publishes a summary and extension of this report in their Replacement Ratio Study&trade;. These replacement rates were calculated by assuming that retirees wish to maintain their pre-retirement lifestyle in retirement. Some adjustments are made for expected changes in spending (e.g. work related expenses will disappear). These replacement rates are dependent on the amount of pre-retirement income and marital status. The main replacement rate tables from the latest Aon Consulting Replacement Ratio Study&trade; are included in this article.

Both the RETIRE and the Aon Consulting reports are very explicit in pointing out that their tabulated replacement ratios are valid for the “average” retiree. They recommend that adjustments always be made when analyzing the retirement needs of specific individuals.

Many popular articles are encountered that recommend replacement rates of 0.75 to 0.85 for retirement planning. Typically the author simply states that these are commonly accepted estimates. Seldom are references supplied to show how such numbers were obtained. The GSU/Aon replacement rates have the advantage of being based on an excellent database (the Bureau of Labor Statistics' Consumer Expenditure Survey) and are derived in a transparent manner.

The GSU/Aon replacement rates are calculated without any consideration as to whether the retiree has the income sources to maintain that level of spending across their entire retirement. In contrast, there are many articles that look at what levels of spending retirees can sustain across their retirement lifespan. These articles also use the term “replacement rate” to express their results. Except in limited circumstances, these two types of replacement rates are not interchangeable. To help avoid confusion, it is suggested that they be thought of as desired replacement rates (e.g. GSU/Aon RETIRE) and achievable replacement rates.

The replacement rate concept
Replacement rates (or ratios) provide a simplified method to estimate your spending needs in retirement. The gross (pre-tax) income you’ll need in retirement is modeled as equaling:


 * Gross Income (retired)  =  Gross Income (pre-retirement)  &times;  Replacement Rate

By rearranging this equation it can be seen that the replacement rate equals a gross income received in retirement divided by a pre-retirement gross income. By gross income, the pre-tax amount is understood.

Replacement rates are expressed in either decimal format or percentage format. These are mathematically equivalent. For example, a replacement rate range of “0.70 to 0.80” is equivalent to a range of “70% to 80%”.

The pre-retirement gross income used is often that received just before retirement. This might be for the year immediately preceding retirement or for an average of a limited number of years prior to retirement (e.g. 5 years). But there is no conceptual reason why a person could not calculate a replacement rate using a gross income received many years before retirement. For retirement planning using the economics-inspired Life Cycle approach, the pre-retirement income is often averaged over an entire working lifetime.

Broadly speaking, there are 2 types of replacement rates encountered in the retirement planning literature. They differ by the type of retirement income used in their calculation: a desired retirement income, or an achievable retirement income. To prevent confusion in the subsequent discussion, these will be referred to as desired replacement rates and achievable replacement rates.

Desired replacement rates
Replacement rates based on a desired retirement income can be used to develop retirement plans that target a specified lifestyle. Most frequently, the specified lifestyle is the one experienced immediately before retirement. In this case the replacement rate being considered is that needed to maintain an unchanged standard of living upon retirement. But a retirement plan (and associated replacement rates) could also target an enhanced or a simplified retirement lifestyle. Such replacement rates implicitly include a tax calculation since it is after-tax funds, not pre-tax income, that are used to supply the desired lifestyle.

For these replacement rates, no a priori consideration is given as to whether the retiree is currently capable of sustaining their desired lifestyle (i.e. desired retirement income) across their anticipated retirement lifespan. A retiree attempting to live at their desired income level could potentially deplete their personal savings prematurely! Nevertheless, desired replacement rates are useful because they can guide a pre-retiree to modify their savings and/or investing approach in order to eventually reach their target retirement lifestyle.

Achievable replacement rates
Replacement rates based on an achievable retirement income typically focus on the specific sources of pre-tax income actually available to the retiree. These types of replacement rates are most frequently encountered in studies of retirement preparedness, where the goal is to better understand the ability of retirees to generate post-retirement income. Some studies focus on the retirement income from just one source, for example just Social Security income or just pension income. When such studies give replacement rates, they refer only to that one source of income relative to pre-retirement gross income. To give a concrete example, it has been reported that Social Security benefits started at age 62 (in 2003) would have replaced 0.31 (31%) of the average recipient’s pre-retirement income. This rate is higher for those with lower pre-retirement income, and vice-versa, reflecting the progressive nature of Social Security's design. It is also possible to add together several or even all sources of retirement income before calculating the corresponding achievable replacement rate. For example, a total replacement rate can be calculated from the sum of annuitized personal savings, Social Security benefits and pension benefits. Notice that replacement rates based on the same pre-retirement income are additive: the total replacement rate is the sum of all individual income stream rates. A special type of achievable replacement rate is derived using an economics-inspired Life Cycle approach. The goal of such an approach is to smooth consumption over both the pre- and post-retirement lifespan. It therefore must produce a lifetime spending and saving plan. Arriving at this lifetime plan requires some rather complicated computer calculations, so specialized software must be employed. The ratio of the resulting smoothed spending rate to some measure of pre-retirement income (for example, the average income across the entire working lifetime) gives the Life Cycle replacement rate. An example of such Life Cycle replacement rates will be presented at the end of this article.

Relationship between desired and achievable replacement rates
The illustration to the right (Fig. 1) will be used to further clarify the relationship between desired and achievable replacement rates. (Note: RR is used as an abbreviation for replacement rate on the vertical axis.)

In an article that deals with achievable replacement rates, Dalirazar et.al show in their Table 1 that such rates increase as the percentage of pre-retirement income saved also increases. This relationship makes perfect sense: the more funds that are saved before retirement, the higher the achievable income that can be maintained after retirement. The dashed line in Fig. 1 illustrates this relationship.

But for desired replacement rates, the relationship to savings is just the opposite. In Aon Consulting's Replacement Ratio Study, Appendix I, the equation for replacement rate will force these rates to drop as savings are increased. The desired replacement rate calculation starts with the gross income immediately before retirement. Taxes and savings are then subtracted to arrive at the pre-retirement spending. The more that is saved, the less that will be left for spending. Aon then assumes that the same spending will be maintained in retirement (i.e. the desired spending upon retirement is the same as just before retirement). Thus as pre-retirement savings are increased, the post-retirement spending is forced to drop. This causes the calculated replacement rate to drop as savings increase. This relationship is shown by the solid line in Fig. 1.

As the savings percentage just before retirement is progressively increased, the achievable RR curve rises and the desired RR curve drops. Eventually they intersect, as shown in Fig. 1. At this savings % the retiree will be able to achieve their desired retirement lifestyle most of the time. However, caution should be used in being too confident in these results, since the achievable replacement rate depends on market returns -- caveat investor.

Miscellaneous replacement rate properties
A total replacement rate can be divided into its individual components (mathematically, it is "additive"). This proves to be extremely useful. It allows both types of replacement rates to be combined in a single retirement plan. For example, suppose that a retiree desires a lifestyle corresponding to an 80% replacement rate. Further, suppose that the retiree's Social Security will only supply a 40% replacement rate. In this case, the retiree's personal savings must be capable of safely supplying the remaining 40% replacement rate (80% total – 40% Social Security = 40% from savings).

Retirement spending is expected to increase over time because of inflation. Therefore the "nominal" replacement rate (nominal retirement income divided by pre-retirement income) would also increase over time. But almost all approaches to retirement planning separate out inflation effects, so that the replacement rates being considered reflect only the relationship between real spending amounts. . In situations where retirement spending categories have strongly different inflation rates, it may prove useful to separate out their real replacement rate components. For example, Steinberg and Lucas propose separating out medical spending from other spending in retirement. This yields two replacement rates: a non-medical rate and a medical rate. This would permit different inflation rates to be applied to each spending component in retirement planning.

Federal and state income taxes have a major impact on desired replacement rates. Changes in the tax code will lead to changes in the size of replacement rates, as well as how these rates vary as a function of gross income level. When taxes in retirement are increased (e.g. on Social Security or pension benefits), the gross income required to supply the desired amount of retirement spending is also increased. This directly increases the corresponding desired replacement rates. Higher income households would tend to be more strongly affected, since low income households typically owe little to no income taxes in retirement. Changes in state income taxes on retirement income should not be overlooked.

In general, achievable replacement rates are not affected by future changes in taxes on retirement income. This is because these replacement rates are not dependent on an implicit tax calculation. But according to Scholz and Seshadri, Life Cycle replacement rates are affected by changes in average effective tax rates.

Replacement rates are most directly connected to pre-tax retirement income rather than after-tax retirement spending. But absolute spending amounts can be derived from them. In fact, the more rigorous approaches to calculating replacement rates estimate absolute spending amounts as an intermediate step. Approaches to calculating spending amounts from replacement rates will be discussed later in this article.

Qualitative influences on desired replacement rates
It is quite common to encounter academic and general-audience articles recommending replacement rates that are less than 1.0 (100%). Factors making a lower retirement income (replacement ratio less than 100%) feasible include the following:
 * Reduction in payroll taxes. Without earned income, most retirees will no longer be paying a 6.2% OASDI or a 1.45% Medicare payroll tax on their salary..
 * Similarly to the disability component of OASDI, private disability insurance is no longer necessary
 * Similarly to the Social Security component of OASDI, the retiree is no longer saving for retirement. The portion of pre-retirement salary that went into savings is no longer needed.
 * Increased federal Standard Deduction. For retirees age 65 or older, the standard deduction is increased by $1,450 for persons filing single, and by $1,150 per person for married filing jointly (in 2011).
 * Partial taxation of Social Security benefits.  Depending on their modified adjusted gross income, Social Security recipients might not owe any taxes on these benefits.  At most, they will be taxed on 85% of these benefits.
 * State income taxes may be lower, because many retirees move after retirement, and many states fully exclude Social Security income from taxes. Other states grant partial tax exemptions for pension and IRA income.
 * The retiree may move, probably to a lower cost-of-living state, and this would decrease their expenses.
 * Potential for higher medical deductibles. With lower taxable incomes, more of the medical expenses of older retirees will potentially exceed the 7.5% limit for itemizing.
 * Mortgage payments may stop. Many retirees finish paying off their mortgage at about the time of retirement, so less income is required.
 * Work-related expenses stop. What had previously been spent on commuting to work, special clothing and meal purchases is no longer required or is needed at a much lower amount.
 * Senior Citizen Discounts. Some ongoing purchases may cost 5-15% less due to special discount programs available to older persons.
 * No dependent children. Many families are no longer spending for the education or living expenses of their children by the time of their retirement.

Not all retirement planners agree that a retiree's replacement rate will be less than 1.0 (100%). There are situations in which a rate of more than 100% might be appropriate. Factors making a higher retirement income (replacement rate more than 100%) feasible include the following:


 * Medical expenses. Medical expenses increase steadily with age for almost all retirees.  Those retiring before becoming Medicare eligible may see a substantial increase in medical insurance premiums until age 65.
 * Travel, vacation and lifestyle changes. Soon after retirement many people increase their spending on special travel and extended vacations that were not feasible while they were still working.
 * Dependents. In some cases retirees may support adult children having physical or mental disabilities.  In other cases the retiree's elderly parents may require partial or complete support, sometimes for expensive long-term health care.
 * Additional services. As retirees age they may need to hire people to help with chores that they can no longer accomplish independently.  Such chores include yard upkeep, snow removal, minor house maintenance, and housekeeping.  Sometimes a retiree's physical impairments may prevent them from driving themselves, requiring them to hire others to do so.

Thinking over the reasons cited for replacement rate reductions as well as increases, it is clear that the heterogeneity of retiree lifestyle and circumstances can easily lead to wide differences in the appropriate replacement rate. One retirement replacement rate doesn't fit all! This will be clearly illustrated in the subsequent discussions.

GSU/Aon RETIRE project
Aon Consulting has teamed with the Georgia State University's Center for Risk Management and Insurance Research to support the RETIRE Project. RETIRE is an acronym for Retiree Income Replacment. The Georgia State University publishes the results as the GSU/Aon RETIRE Project Report ; Aon Consulting publishes its summary and extension of the findings as the Replacement Ratio Study&trade;. The latest studies, released in 2008, are the seventh update in the series. The primary question targeted by this study series is, "How much income will I need at retirement to maintain my standard of living?" Thus it must be understood that the replacement ratios being envisioned by this study assume no changes in lifestyle upon retirement. These are desired income replacement rates.

Methodology
The replacement ratios estimated in the 2008 study are derived primarily from the Bureau of Labor Statistics' Consumer Expenditure Survey for 2003, 2004 and 2005. The replacement rates were calculated based on data for 12,823 "working" households and 6,498 "retired" households.

The baseline calculation is for a married couple with one wage earner who retires at age 65 (i.e. when Medicare eligibility begins). Their non-working spouse is age 62. The calculation starts with an average, pre-retirement gross (pre-tax) income. From this income are subtracted average taxes and retirement savings. This gives the average pre-retirement spending. This is then converted into after-retirement spending by either adding or subtracting retirement-related average changes in spending. Finally, post-retirement taxes are estimated and added to the average spending to arrive at an average post-retirement income. The final replacement ratio is the post-retirement income divided by the pre-retirement income. These calculations are shown in detail in Appendix II and III of the latest study.

GSU/Aon replacement ratios
The baseline replacement ratios for pre-retirement incomes ranging from $20,000 to $250,000 are shown in the "Total Ratio" column of the following table. The average Social Security benefit for each family, expressed as a percentage of pre-retirement income, is also shown in the table. The middle column, "From Private and Employer Sources", is the difference between the other two column entries (i.e. it is derived, not directly calculated).

'''Table 1. Baseline Replacement Ratios (married ages 65/62, one working)'''

Significant points that stand out in this table are:


 * Total replacement rates are highest for the low income households, drop to a minimum for families with pre-retirement incomes of $70,000 to $80,000, and then rise again for extremely high income households.
 * Social Security benefits supply a much higher portion of the retirement income stream for lower income households than for higher income households.
 * Conversely, higher income families looking to maintain their pre-retirement lifestyle in retirement will need to supply a correspondingly higher percentage of their income via pensions or private savings withdrawals.
 * In general, high-income individuals can maintain their pre-retirement lifestyle only by having substantial personal savings. Income from Social Security and employer pensions will typically not be nearly enough to maintain their lifestyle.

Besides the baseline scenario, the Aon study also calculates replacement ratios for other household types. The four types of households considered are:


 * Baseline - a married couple with one wage earner who retires at age 65. Their non-working spouse is age 62.
 * Single person age 65
 * Married couple with one wage earner who retires at age 65. Their non-working spouse is also age 65.
 * Married couple ages 65 and 62. Both earn wages.

The following table shows the replacement ratios for all four scenarios.

Significant points that stand out in this table are:


 * The replacement rates for singles show less variation across income levels than for any of the married scenarios.
 * The replacement rates show minimal variations between the 3 married scenarios. Below the $60,000 pre-retirement income level there are no differences.  This is supposedly because such retirees do not pay significant income taxes.

Adjustments to replacement ratios
The RETIRE and Aon Consulting reports are very explicit in pointing out that their replacement ratios (shown in the tables above) are valid for the "average" retiree. They recommend that adjustments be made when analyzing the retirement needs of specific individuals. Factors that might lead to such replacement ratio adjustments include the following:


 * Individual differences in savings rates. The calculation approach used in the GSU/Aon RETIRE study assumes an "average" savings rate in the years just before retirement.  If a specific individual or family saves more than average, it means that their preferred lifestyle can be maintained by spending less than average.  Their replacement ratio should therefore be lowered to adjust for their higher saving rate.  Conversely, individuals or families saving less than average would need to have their replacement rate adjusted upward.
 * Differences in types of savings. Withdrawals from tax-deferred retirement savings accounts (e.g. traditional IRA or 401(k) accounts) are taxed, leading to a larger desired replacement rate.  But withdrawals from Roth IRAs or brokerage (taxable) savings do not increase taxable income and therefore lead to a lower replacement rate.
 * Changes in big-ticket spending items. Effective replacement ratios can be reduced for individuals who retire immediately after paying off their mortgage or immediately after finishing payments for a child's college expenses.  Once again, their pre-retirement spending lifestyle was at a lower than average level.
 * Specifics of retiree medical coverage. The Aon study's baseline calculation assumes that medical expenses increase by somewhere between $1,000 to $1,500 upon retirement.  But in some cases medical expenses might increase by a significantly larger amount.  For example, retirement before Medicare eligibility will lead to much larger increases in medical insurance premiums.  A similar effect can arise when people who had employer-subsidized medical insurance while working lose that support after retirement.

Medical expense changes at retirement can have a particularly strong effect on the replacement ratio. The Aon study illustrates this by recalculating these ratios under three different medical expense scenarios:


 * Best Case: essentially no change in medical expenses upon retirement.
 * Baseline Case: medical expenses are assumed to increase by about $1,000 to $1,500 annually, depending on income.
 * Worst Case: medical expenses are assumed to increase by $400 per month upon retirement. This $400 represents the approximate monthly costs of Medicare part B and D premiums, plus a premium for supplemental ("Medigap") coverage.

''' Table 3. Total Replacement Ratio (married, ages 65/62, one working)'''

As this table shows, the total replacement ratio for lower income level families can be substantially shifted by their actual retirement medical expenses.

If your individual spending situation is different from the averages used in the Aon study, you might consider working with an experienced retirement advisor to adjust the tabulated replacement ratios. Alternatively, you could imitate the Aon study approach and calculate personalized replacement ratios from scratch.

Replacement rate changes with aging
Although often ignored, there are reasons to expect replacement rates to drop modestly over the course of retirement. Various Studies of retirement spending have shown that real spending in retirement typically drops steadily until at least the late 70's. For example Fisher et.al. analyzed the BLS Consumer Expenditure data and determined a "real dollar" median spending drop at retirement of 2.5%, followed by about a 1% per year median spending drop over the next 15 years. Such a spending drop would also cause replacement rates to decrease over time. At even older ages spending and associated replacement rates might continue dropping if the retiree has adequate Medigap and long-term care insurance. But without adequate medical insurance, end-of-life medical costs have the potential to jump significantly, increasing the corresponding replacement rates.

Tacchino has proposed two methods to factor in spending and replacement rate reductions with aging. If it is necessary to use one replacement rate for the entire retirement time period, then a weighted-average of the various replacement rates could be calculated and utilized. An alternative approach would be to use one replacement rate before age 75 and a second, lower rate afterwards.

Life Cycle replacement rates
The RETIRE and Aon Consulting studies showed how desired replacement rates can change depending on gross income, marital status and medical spending assumptions. In this section the dependency on measurement time period is briefly explored.

Desired replacement rates in the GSU/Aon RETIRE study were calculated assuming an equivalent consumption lifestyle in the immediate time period before and after retirement. If the time period of equivalent consumption is broadened, the replacement rates should change. If this time period is expanded to cover the entire working and retirement lifespan, then it becomes identical to the Life Cycle approach.

From this perspective, replacement rates calculated using the Life Cycle approach constitute one extreme in a continuum of possible desired replacement rates. The GSU/Aon replacement rates constitute the other extreme. It is instructive therefore to compare a set of Life Cycle replacement rates to those published by Aon Consulting. The Life Cycle replacement rates published by Scholz and Seshadri will be used for the comparison. These are shown in the following table (Table 4).

'''Table 4. Life Cycle Median Replacement Rates (RR)'''

The first column in Table 4 categorizes households into income deciles: higher deciles represent higher gross incomes. They chose to use two different pre-retirement gross incomes for the replacement rate calculations. The "Lifetime Income" column uses the real household income averaged over the entire working lifetime as the measure of pre-retirement gross income. The "Top 5 Earning Years" column uses the real household income averaged over the ninth through the fifth years before retirement (i.e. the years during which household income is usually maximized). Replacement rates calculated using the "Top 5 Earning Years" income are smaller, since for all income deciles except the lowest this is a larger average income.

Scholz and Seshadri did not give dollar amounts for each of their decile ranges in Table 4. But they did supply dollar amounts for the median (50%) incomes: about $35,000 for the Lifetime Income, and about $38,000 for the "Top 5 Years" income, both in 2004 dollars. From Table 1 above, the GSU/Aon median replacement rate for a $40,000 income is 0.85. This is substantially larger than either 50% decile Life Cycle replacement rate estimated from Table 4 (0.63 and 0.55). It is evident that the GSU/Aon replacement rates are systematically higher than the Life Cycle rates. As expected then, replacement rates also change depending on the time period covering the lifestyle equivalency.

On the whole, the GSU/Aon replacement rates are systematically higher than the Life Cycle rates. This indicates that the typical retiree tends to have a higher consumption lifestyle around the time period of retirement compared to earlier or later in life. After all, if they tended to exhibit a steady consumption across their entire life, their replacement rates would be lower and closer to the Life Cycle approach values. This conclusion is consistent with the observation that real spending for the typical retiree drops steadily as they age. Looked at another way, this indicates that, on average, people tend to spend proportionally more on themselves during their maximum earning years just before retirement.

Calculating spending from replacement rates
In some situations it may be useful to estimate your after-tax spending in retirement. This can be derived from your gross retirement income by subtracting federal and state taxes. The following two equations, when combined, show the connection between spending and replacement rates.


 * Gross Income (retired)  =  Gross Income (pre-retirement)  &times;  Replacement Rate
 * Spending (retired)  =  Gross Income (retired)  &minus;  Federal taxes  &minus;  State taxes

The first equation shows how the gross income in retirement is estimated by means of a replacement rate and a pre-retirement gross income. The second equation shows how retirement spending is estimated from the gross income in retirement by subtracting taxes. Estimating your federal and state taxes will require you to identify the specific sources of income that combine to equal your gross retirement income. This is required because these different sources of income often are taxed at different rates. Typical sources of retiree income include Social Security, pensions, interest, dividends, capital gains, withdrawals from tax-advantaged savings (e.g. 401(k)), 403(b), and traditional IRAs), and withdrawals of tax-free savings (Roth IRAs, principal in bank or brokerage accounts).

The easiest method of calculating the federal and state taxes is using commercial software. The only drawback is that such software assumes the current tax rate structure. What if you anticipate future tax rate increases and want to incorporate this anticipation into projected retirement spending? In this case you will need to either use software or a tax spreadsheet that allows you to change the tax rate assumptions, or perform the tax calculation by hand.