How Much Do We Need To Retire?

The growing number of retirees, their increased life expectancies and the size of their savings led us to wonder: How much accumulated income-producing wealth is needed to maintain a desired lifestyle? In other words, given a target level of spending, how much did a couple need to have saved, in either tax-deferred retirement accounts or other financial assets, to feel confident that they would survive

The growing number of retirees, their increased life expectancies and the size of their savings led us to wonder: How much accumulated income-producing wealth is needed to maintain a desired lifestyle? In other words, given a target level of spending, how much did a couple need to have saved, in either tax-deferred retirement accounts or other financial assets, to feel confident that they would survive until age 90 with their accustomed level of spending?

To find the answer, we first have to account for the difference between:

  1. qualified plan balances or IRAs,1 all of which constitute untaxed income until withdrawn and are subject to the minimum distribution requirements (MDRs) contained in the Internal Revenue Code and the Internal Revenue Service's regulations; and
  2. post-tax savings. Beginning in 2010, the IRC will allow taxpayers to convert part or all of their IRAs into Roth IRAs, that being the only means of accumulating a large enough fund out of which a retirement could be financially sustained.2

Next, we have to factor in likely investment performance, which would have as much effect on the ability to meet spending goals over the long haul as the initial balances in the three types of funds.

Looking at IRAs, the MDR affects the behavior of retirement savings in at least three ways:

  1. It may result in rates of withdrawal from the accounts that are higher than financially prudent.
  2. Whatever is withdrawn will generate income taxes on the entire annual distribution, lowering its value to the recipients by the taxes paid.
  3. As illustrated in 2008, the MDR operates independently of investment results so that one cannot cut back the rate of withdrawal in a year of poor investment returns. Those funds will not be available to rebuild the account when the investment climate becomes more favorable.

By comparison, if we had the equivalent sum in post-tax savings or a Roth IRA, we could set an annual withdrawal rate that virtually insures the retirees would continue to have adequate resources and, to the extent the principal was invaded, that portion of the withdrawal would not itself generate any income tax. Or we could moderate spending for that year and defer consumption. On the other hand, given the deductibility of contributions to qualified plans and IRAs during the years a client is earning income, accumulating tax-deferred retirement savings is cheaper and easier than doing the same with after-tax dollars.

We anticipate that the pattern of spending probably would change over time, with expenses like travel and consumption highest in the early retirement years and those like health care expenses highest in later years. So we posit an unaltered spending rate throughout retirement on the not unreasonable assumption that, one way or another, a retired couple will find a way to spend their income. We also assume the solvency of the Social Security system for the next 20 years and an average annual inflation rate of 3 percent.3

Issues for IRAs

What we find when we look at IRAs is trouble ahead. (See “What's Required,” this page.) MDRs for the early years may not provide enough money to meet desired lifestyle needs. This shortfall naturally would encourage the retirees to withdraw more, creating a potential problem for them down the road. Also, and a little more frightening in the current economic climate, the IRA would have to experience solid investment performance to hold its own and allow for distributions to keep pace with inflation.

Most, if not all, of the software readily available to make projections uses an average rate of return style of analysis that tends to understate or eliminate these problems, depending on the rate selected. Thus, if one posits 10 percent growth and a 5 percent rate of spending or withdrawal over 10 years, the average rate of return model indicates no problem as the IRA continues to grow. The cushion built in the early years buffers the impact of larger percentage withdrawals in later years, and the model posits that everything works out just fine.

Unfortunately, these models have two flaws. As we all know too well, financial markets do not behave in the linear fashion posited by the average rate of return model, nor do an individual's investments. Equally important: if we assume that the IRA will have to provide income, the path of actual returns will produce widely different results, even with the same average rate of return.

The longer the period for the projection, the worse these methodological defects grow. Just look at a 20-year window, assuming retirement at age 70 and the death of the surviving spouse at age 90. If we assume a 6 percent rate of return, the IRA will outperform the MDR withdrawals for the first 10 years or so. But then the IRA will only start to shrink when one reaches age 83.

The actual path of returns produces very different results. Let's posit an example to dramatize the differences. Assume we have 2 clients who each have a starting balance of $5 million, and who constantly withdraw $250,000 per year (we'll disregard inflation). If we use a 10 percent average rate of return, the actual returns of the S&P 500 from 1993 to 2002 (Path 1), then reverse their chronology (Path 2), we see the variation that can occur between the two clients. (See “Real Life Is Unpredictable,” this page.) In either case, the average rate of return came out, of course, to 10 percent.

At the end of the eight years, the client who followed Path 1 yields an account that is 50 percent higher than Path 2. The account in Path 1 is 80 percent higher than at the outset. In Path 2, it's 20 percent higher. If the accounts were not used to meet spending needs, the two paths would have converged. But that is not how retirement assets are used or the purpose for which they're intended. Of course, the same thing would happen if the scale of the account and the spending were reduced. The same results would occur if one used a $1 million account and a $50,000 spending rate.

The two paths also illustrate the effect of positive and negative compounding. The dollars withdrawn from the account in the earlier years along Path 2 are not available to help recoup the investment losses in the later years. Conversely, the excellent returns in Path 1 during the early years soften the impact of the negative results in the later years.

This simple example illustrates the fallacy of the average rate of the prevailing return model. That model may give a working individual an approximation of the size of his account 20 years down the road, but only if there were no withdrawals. To the extent there are shortfalls, these can be remedied by different investments or higher contributions. But there ends the utility of the constant rate of return model.

The approach that more nearly approximates reality is the Monte Carlo simulation. The Monte Carlo approach reflects the uncertainties about how the financial markets will perform by using random numbers to generate a path of returns. But in this context, the numbers are not utterly random. Inflation rates do not vary between 100 percent in Year 1 to minus 100 percent in the following year; nor do interest rates fluctuate wildly from one extreme to another from year to year. Instead, the range of actual outcomes reflects the historical, observed volatility in different capital markets or types of investments. How each type of investment fares from year to year usually stays within the range of volatility, but the returns may be randomly positive or negative within that range. For example, if interest rates historically fluctuate by at most 2 percent from year to year, then if the 10-year Treasury note yields 8 percent in Year 1, it will yield anywhere from 6 percent to 10 percent in Year 2.

This second model also reflects the historical correlations between different investment vehicles and critical factors such as interest rates and inflation. As we know from experience, high U.S. inflation rates and correspondingly high interest rates do not produce an environment in which domestic stocks perform well. They might, but the volatility constraints limit the upside potential of U.S. equities. Other types of investments, such as international stocks, do not exhibit the same degree of correlation to U.S. interest rates and inflation and can appreciate significantly even if domestic equities are falling. We assumed that each of the retirement options was invested in a geographically diversified portfolio, but not in alternative investments that also carry a low degree of correlation with the domestic stock market but at considerably more risk.

Each of the various scenarios described in this article is run through 10,000 trials or simulations. From those results, one can see the likelihood of various outcomes with the focus on results with a 10 percent to 90 percent probability. As one would expect, one arrives at a 50 percent level of confidence within the levels where one finds the largest cluster of results.

For obvious reasons, people would like to see their level of spending and their lifestyles continue uninterrupted by retirement and the consequent end of earned income. As a result, we used fairly large retirement accounts in our analyses and assumed that the retired couple wanted to be able to spend $200,000, adjusted for inflation, after paying taxes on their withdrawal from their retirement accounts, or in largely tax-free or totally tax-free income from the post-tax savings and the Roth IRA accounts. For simplicity's sake, we treated the retirement assets as one large IRA or account. Again, these results are scalable, and one can decrease the size of the accounts and the amount of spending proportionately. We also assumed that retirement would last 20 years, that is, one of the retirees would live to be 90. The odds of that occurring are now one in four, and anyone planning for retirement needs to take current longevity into account.

As a retiree ages and life expectancy declines under the IRS' uniform table of life expectancy, the rate at which funds must be withdrawn increases. We took this into account, assuming that in years in which the required distributions exceeded spending needs, the surplus was invested in a taxable investment account whose returns were subjected to the same Monte Carlo simulation process. In good investment years, the retirees were building a post-tax account to act as a reserve against years with lower returns or higher MDR requirements. Keeping a reserve guards against the unpredictable nature of medical expenses. We expected and found that a larger accumulation at the outset increased the odds that such a taxable investment account would ever exist. (See “Option 1: IRA, The Nature of Retirement Savings,” p. 23.)

How Much is Needed?

These results are not surprising. Given the income taxes imposed on every withdrawal, having $200,000 to spend means taking about 10 percent annually of the starting value of a $3 million IRA and roughly 6 percent of a $5 million IRA. As a corollary, if one looks at the value of the taxable account created by excess distributions, it will be zero for an IRA of $3-to-$5 million and have a 50 percent chance of being worth $1.2 million in the 20th year after retirement for a $6 million IRA. In other words, the larger the IRA, as one would expect, the higher the probability becomes of building an additional reserve out of post-tax retirement income.

The surprising result of this set of simulations is that the apparently “safest” asset allocation carries the highest risk that the account will decline in value to an uncomfortably low level and only a small advantage in terms of diminishing the risk of depletion. The 40/60 allocation between stocks and bonds holds a slight advantage over the 60/40 allocation in terms of risk of depletion, but the more aggressive strategy holds, as one would expect, a higher mid-point and top. Either of these investment approaches underlines the need for growth in the IRA if it is to serve its purpose.


We then looked at what happened if one did not accumulate the IRA size required to live in comparative serenity while maintaining the level of spending. The obvious solution is to reduce spending and fit one's budget to one's means. (See “Adjust Accordingly,” this page.)

In general, we found that reducing spending by $30,000 for each $1 million shortfall resulted in the necessary security that the retirement account would survive for at least 20 years. Bear in mind that spending means after-tax dollars and that one's tax rate may decline as one withdraws less from the account. Obviously, the lower the spending, the higher the chances are that one will be able to maintain the initial level and not have to cut back on one's lifestyle.6

Returning to our discussion of “Real Life Is Unpredictable,” p. 22, it would require nerves of steel if one experiences Path 2, as retirees in 2000 did, and not take some remedial action. The fact that it will all come out in the wash 50 to 90 percent of the time does not lend much comfort if one is convinced by three years' experience that one is living through the bottom 10 percent of all outcomes. As that experience shows, this study casts further doubt on the advisability of relying solely on qualified plans as the means of funding retirement. Ideally, one should build up a taxable stock of savings to supplement the qualified plan and to provide a mechanism for setting the rate of spending or withdrawal that works for the retirees and does not have to satisfy any governmental mandates.

Bottom Line

Given the current level of savings and the expectations that many have that they will live as well in retirement as they did while in the work force, it may come as a shock to discover how large one's savings need to be to achieve that goal. As a corollary, one must either save more than previously thought or re-examine the premise that disposable income will remain as freely available after retirement as it was before.

The ability to use tools more sophisticated than the average rate of return model also underscores the complexity of retirement planning. Each individual needs to address the related issues of spending and saving as early as possible so that his accounts — qualified and taxable — can increase to a level that holds out a realistic chance of maintaining his standard of living, however long he may live. Alternatively, one must analyze spending needs after retirement if savings will not allow the luxury of spending as though retirement had not occurred. Whether this leads to a decision to defer retirement until one can attain one's financial goals or curtail post-retirement outlays will depend on one's value system. However one looks at the problem, it has more wrinkles than one would have thought.

It also mandates a higher level of savings at an earlier age than suggested by the prevalent wisdom. The emphasis on the IRA and its counterparts represents a laudable effort to make that process seem as painless as possible during the accumulation process and maximizing contributions to a qualified plan makes sense, certainly for the younger participant. In the long run, unless one can accumulate a large IRA during the working years, there is no substitute for saving out of post-tax income for the proverbial rainy day. Unlike the other alternatives, that money belongs entirely to the owner.

— The authors wish to acknowledge the data from the Wealth Forecasting Analysis (SM) provided by AllianceBernstein's Wealth Management Group; the conclusions and observations made by the authors do not reflect those of AllianceBernstein, which was not involved in the authorship of this article.


  1. For ease of reference, we have referred to all of such accounts as IRAs, all of them being subject to the same minimum distribution requirement (MDR) rules.
  2. See Internal Revenue Code Section 408A(d)(3).
  3. In selecting this rate of inflation, we are mindful of the fact that the rate of inflation for those over the age of 65 differs from the consumer price index (CPI), in large part because of the higher component of health care spending among that population and the fact that health care costs inflate more quickly than the CPI. However, for purposes of looking at investment results, the general rate of inflation influences financial markets rather than the rate borne by the elderly.
  4. This model reaches the same conclusion as Natalie Choate did in her seminal Life and Death Planning for Retirement Benefits (6th edition 2006), ATAXPLAN Publication, at pp. 270-71.
  5. Ibid. at p. 278.
  6. One more tool for the retiree is the Health Savings Account (HSA) described in Section 223 of the Code. Given the dollar limitations, the end of eligibility to make contributions after age 65, and the requirement that, to remain tax-free, expenditures must be for “qualified medical expenses,” the utility of the HSA lies primarily in its ability to help a retiree fund long-term care insurance or medical expenses, thereby eliminating one more source of concern to all of us as we age.

Deborah V. Abildsoe is managing member of Asset & Retirement Investment Associates LLC in Guilford, Conn. Irving S. Schloss is a partner at Murtha Cullina LLP in its New Haven, Conn., and Madison, Conn. offices.


Government mandates specific percentages be pulled from IRAs over the years. Those “distributions” then are taxed according to a client's income bracket

Years of Age Minimum Distribution
Life Expectancy Requirement (Percentage) 40 Percent 30 Percent
70 27.4 3.65% 2.19% 2.55%
71 26.5 3.77 2.26 2.64
72 25.6 3.91 2.34 2.73
73 24.7 4.05 2.43 2.83
74 23.8 4.20 2.52 2.94
75 22.9 4.37 2.62 3.06
76 22 4.55 2.73 3.18
77 21.2 4.72 2.83 3.30
78 20.3 4.93 2.96 3.45
79 19.5 5.13 3.08 3.59
80 18.7 5.35 3.21 3.74
81 17.9 5.59 3.35 3.91
82 17.1 5.85 3.51 4.09
83 16.3 6.13 3.68 4.29
84 15.5 6.45 3.87 4.52
85 14.8 6.76 4.05 4.73
86 14.1 7.09 4.26 4.96
87 13.4 7.46 4.48 5.22
88 12.7 7.87 4.72 5.51
89 12 8.33 5.00 5.83
90 11.4 8.77 5.26 6.14