It's well known that a traditional IRA is a great way to save. Assets in an IRA can grow tax-deferred for decades. But tricky planning issues arise when a married couple has combined assets in excess of the estate tax exemption (currently, $3.5 million)1, but one or both of the spouses have insufficient assets outside of an IRA to fund a credit shelter trust (CST) fully. When that's the case, many estate planners question whether clients should use their IRAs to fund a CST at death, because doing so generally means giving up a lot of potential income tax deferral, and often taxing the IRA distributions at higher rates.

Our research shows that wasting the estate tax exemption of the first spouse to die by eschewing the CST creates an even greater tax burden — and is likely to be a mistake.

IRA in a CST

A CST is a fundamental part of the estate plans of most wealthy married couples. Typically, upon the death of the spouse who dies first, the plan puts the estate tax exemption amount into a CST that continues during the surviving spouse's lifetime. Often, the trustee may (but need not) make distributions to the spouse or the couple's descendants. Regardless, the CST is not subject to estate taxes at the surviving spouse's death. So, any assets remaining in the trust can pass to the couple's descendants free of estate taxes.

The two most important potential drawbacks of funding a CST with an IRA are:

  • Loss of deferral: The required minimum distributions (RMDs) from an IRA payable to a CST generally are significantly larger than they would be if the IRA were payable outright to the surviving spouse (who could roll over the IRA). In a worst-case scenario, the IRA must be fully distributed by the end of the fifth year after the year in which the IRA owner dies. In any event, there's usually a substantial loss of income tax deferral.
  • Higher tax rates: Undistributed trust income may be subject to high income tax rates: For 2009, a trust reaches its highest income tax bracket at about $11,000 of taxable income — compared with about $373,000 for individuals. For a CST that accumulates rather than distributes some or all amounts it receives from an IRA, this is clearly a negative.

As a result, some planners conclude that it's always better simply to abandon fully funding the CST and to make the IRA payable to the surviving spouse. But by stress-testing several solutions, we can better understand the range of wealth outcomes associated with various planning strategies, and the overall dollar savings that can be realized.

The Options

The question boils down to whether it's more advantageous to minimize estate taxes or minimize income taxes. To quantify the trade-offs involved, we looked at three distribution options for the IRA at the account owner's death.2

  • In the first option, the IRA owner abandons funding the CST. He simply designates his spouse as the beneficiary of his IRA, who rolls it over into an IRA in her own name and designates the couple's child as the beneficiary.3 That's the simplest strategy and results in the smallest RMDs. During the surviving spouse's lifetime, RMDs are based on the IRS' “Uniform Lifetime Table:” Each year the RMD is determined by the joint life expectancy of the spouse and a hypothetical person 10 years younger.4 After the spouse's death, RMDs are based on the child's life expectancy in the year after the year of the spouse's death, progressively reduced by one in each subsequent year.5 But recall that the first spouse to die has completely wasted his or her estate tax exemption.

Continue on Page 2

  • Another option is to have the decedent use his IRA to fund a CST structured as what's often called a “conduit trust.”6 With the conduit trust, the trustee must immediately distribute to the surviving spouse any amount paid from the IRA to the CST. Although RMDs from a conduit trust are larger than under the first option outlined, the rules still are fairly favorable: During the surviving spouse's life, the RMD is recalculated annually based on the spouse's life expectancy in that year. After the spouse's death, RMDs are based on the life expectancy of a person of the spouse's age at her death, progressively reduced by one in each subsequent year.7 Distributions to a conduit trust “pass through” to the surviving spouse, potentially avoiding the higher trust income tax rates. Most importantly, a conduit trust takes advantage of the estate tax exemption of the first to die. But because distributions from the IRA are moved into the surviving spouse's estate, only what's left in the IRA at the spouse's death escapes estate taxes.
  • A third option is to simply abandon income tax deferral to maximize estate-tax savings by making the IRA payable to what's sometimes called a “non-see-through” trust, that is to say, a trust no beneficiary of which is treated as a designated beneficiary of the IRA. With a non-see-through trust, at the trustee's discretion, the RMDs can all stay in the CST, increasing the amount protected from estate taxes. The catch is that if the IRA owner dies before his required beginning date (RBD), the entire IRA must be distributed by the end of the fifth year after his death.8 This ends further income tax deferral and taxes the distribution at trust income tax rates.9 (See “Three Ways To Save,” p. 20.)

To determine which option is best, we stress tested each by running a mortality adjusted analysis on our wealth forecasting system, which uses a Monte Carlo model that simulates 10,000 plausible future paths of return for each asset class and inflation and produces a probability distribution of outcomes. This model does not randomly draw from a set of historical returns to produce estimates for the future. Instead, our forecasts:

  1. are based on the building blocks of asset returns, such as inflation, yields, yield spreads, stock earnings and price multiples;
  2. incorporate the linkages that exist among the returns of various asset classes;
  3. take into account current market conditions at the beginning of an analysis; and
  4. factor in a reasonable degree of randomness and unpredictability.

We looked at the following scenario: A 70-year-old husband dies before his RBD. He's survived by his wife of the same age and a 45-year-old son. The husband's total assets of $3.5 million are in an IRA allocated in a 60 percent/40 percent stock-bond mix. The wife is spending $250,000 a year after taxes, growing with inflation, and has $5 million in her own taxable account (which is also allocated 60/40). We assume that the surviving spouse's income is taxed at a 28 percent rate (31 percent beginning in 2011) and that income taxed to the CST is taxed at a 35 percent rate (39.6 percent beginning in 2011). We also assume that distributions are made from the CST to the spouse only after the spouse exhausts her taxable account (except in the case of the conduit trust, which must distribute IRA withdrawals to the spouse).

So, which option is likely to provide the couple's child with the most after-tax wealth?

Our analysis shows that forgoing the estate tax savings of a CST is highly likely to be a mistake. A simple spousal rollover fared the worst of the three options in the upside, downside and median cases.

Put simply, it's highly unlikely that the additional income tax deferral from a rollover will make up for the incremental estate taxes that will accrue from wasting the IRA owner's estate tax exemption. Based on our estimates of the child's inflation adjusted after tax wealth 20 years after the death of the surviving spouse, the probabilities tell the story well.10 (See “And the Winner Is …” p. 21.)

Interestingly, option three, the non-see-through trust — which results in a distribution of the entire IRA five years after the account owner's death — leaves the couple's child with more after-tax wealth than either a spousal rollover or a conduit trust more than two-thirds of the time. In the median case, it provides the couple's child with almost $1 million more of inflation-adjusted wealth than the conduit trust and almost $2 million more than the spousal rollover.

Continue on Page 3

Way To Go

The moral of the story? As we see it, for most married couples with combined estates large enough to be subject to estate tax, there are two:

  1. If an IRA must be used to fully fund a credit shelter trust, it should be used; and
  2. The credit shelter trust should not be structured as a conduit trust.

Bernstein Global Wealth Management is a unit of AllianceBernstein L.P. Bernstein does not offer tax, legal, or accounting advice. In considering this material, you should discuss your individual circumstances with professionals in those areas before making any decisions.


  1. Internal Revenue Code Section 2010(c).
  2. In each case, we assume that the IRA owner dies before his required beginning date (RBD). See Treasury Regulations Section 1.408-8, A-3.
  3. Treas. Regs. Section 1.408-8, A-5(a).
  4. Treas. Regs. Section 1.401(a)(9)-9, A-2.
  5. Treas. Regs. Section 1.401(a)(9)-9, A-5(c)(1).
  6. See, for example, Natalie B. Choate, Life and Death Planning for Retirement Benefits, Ataxplan Publications, Boston, p. 314 (6th ed. 2006).
  7. See Treas. Regs. Section 1.401(a)(9)-4, A-5(c)(2).
  8. See Treas. Regs. Section 1.401(a)(9)-4, A-5(c)(3).
  9. A fourth possibility — and the best option of all in our analysis — is a “non-conduit” trust under which the surviving spouse is treated as the designated beneficiary of the IRA, which would allow RMDs to be based on the surviving spouse's life expectancy in the year after the year the owner dies, reduced by one in each subsequent year. Treas. Regs. Section 1.401(a)(9)-4, A-5 (c)(1). The problem is that the applicable Treasury regulations are highly ambiguous regarding how to structure a trust that is not a conduit trust and that is not subject to the five-year rule. See Treas. Regs. Section 1.401(a)(9)-5, A-7; see also Choate, supra note 6, at pp. 312-14.
  10. In our mortality-adjusted analyses, the lifespan of an individual varies in each of the 10,000 trials run by our wealth forecasting system in accordance with the mortality tables. For example, there's a 50 percent chance that a 70-year-old woman will survive until age 88. Consequently, in a mortality-adjusted analysis for that woman, she will survive until at least age 88 in half of the 10,000 scenarios we model. In the other half, she will die at a younger age. Mortality simulations are based on the Society of Actuaries, Retirement Plan Experience Committee Mortality Tables RP-2000.

David L. Weinreb (left) is a director and Gregory D. Singer is director of research in the Wealth Management Group at Bernstein Global Wealth Management in New York