The sequence, and then the arithmetic that decides whether it was worth doing.
The grantor funds an irrevocable trust with property expected to appreciate, most often concentrated stock or an interest in a closely held business.
The instrument fixes the annuity and the term. The annuity is a stated dollar amount or a fixed percentage of the initial value, payable at least annually, and may step up by no more than 120 percent of the prior year's payment.
The gift is computed once. The retained annuity is valued under section 7520 using the rate the IRS publishes for the month of the transfer, and the remainder, meaning the transferred value minus the annuity's value, is the reportable gift. Where the annuity is sized so that its value nearly equals the property transferred, the reportable gift is close to zero, which is what practitioners mean by a "zeroed-out" GRAT.
The trustee pays the annuity every year, in cash or by distributing property in kind at its value on the payment date.
The term ends. Whatever is left passes to the remainder beneficiaries or to a trust for them, with no further transfer tax.
A hypothetical, with the arithmetic set out. Priya transfers $1,000,000 of stock to a two-year GRAT and retains an annuity of $515,000 a year, an amount chosen so that the value of what she keeps is close to the value of what she put in and the reportable gift is nearly nothing. The dollar figures here are illustrative; the annuity that produces that result depends on the assumed rate for the month, which changes.
Suppose the stock returns 12 percent a year. Year one: $1,000,000 grows to $1,120,000, the trust pays her $515,000, and $605,000 remains. Year two: $605,000 grows to $677,600, the trust pays her $515,000, and $162,600 passes to her children, having used almost none of her lifetime exclusion.
Now suppose the stock returns 2 percent instead. Year one: $1,020,000 minus $515,000 leaves $505,000. Year two: $515,100 minus $515,000 leaves $100. Priya has her money back, her children have essentially nothing, and the cost of the attempt was the drafting and administration. That asymmetry, a real payoff when the property outperforms and a near-zero cost when it does not, is the reason the structure is usually built with a short term and repeated rather than run once over a long one.