Every few years, it seems like the marketplace produces a "study" of variable annuities that makes the product look too good to be true—a creative fusion of research and PR. Now it looks like a company I've always respected—Ibbotson Associates—has joined this unfortunate litany.

I'm referring to a white paper titled "Retirement Portfolio and Variable Annuity with Guaranteed Minimum Withdrawal Benefit (VA+GMWB)." Published in October 2007, the paper evaluates the guaranteed minimum withdrawal benefits that are routinely sold as VA riders. This objective evaluation was (as the cover page discloses) sponsored by Nationwide Financial, whose Best of America annuities are among the hottest-selling products in the industry. The company, a footnote on page three says, was helpful in this analysis, even suggesting the paper's hypothesis.

You don't have to read far to find it: "We have developed a hypothesis that the GMWB will help improve the overall retirement income levels without increasing income risk levels" (p. 3); "variable annuities and payout annuities...can help investors hedge market risk and retirement income risk" (p. 5). "Our hypothesis is that adding VA+GMWB to traditional portfolios will improve the overall retirement income levels without increasing income risk levels" (p. 10).

For the benefit of readers who do not sell annuities for a living, the paper offers a quick explanation of how the GMWB works. When you buy an annuity with this rider, you can withdraw 5% of the cash value each year. Or, if the market has dropped, you can look back to the cash value on last year's anniversary date, which becomes the "benefit base," and take 5% of that higher figure. If the markets tank for five straight years, you keep taking this same 5% of the same benefit base until your portfolio climbs out of the hole.

You can see how this works in the white paper's Table 1, which is essentially a spreadsheet showing an annuity's cash values and distributions from 1979 through the end of 2006. The GWMB really kicks in after the down market of 2000, when the account value drops from a 1999 high of $3.8 million to, in successive years, $3.4 million, $2.8 million and $2.2 million at the end of 2002. Even though the bear market is reducing the portfolio each year, even though it never gets back up to that 1999 high, the contract keeps paying what it paid in 1999: $191,578.

Who Really Pays?

It's helpful to note a few things that aren't made clear in the white paper. First, this account isn't annuitized. Although the distribution is being held constant, the insurance company that sold the rider isn't reaching into its pocket to pay the difference. All the distribution money is still coming from the annuity holder's (diminishing) investment account—ultimately from the pockets of the people who will inherit it. The only way the insurance company would have to make a payment—and justify the rider cost—is if the account vanished all together, forcing the insurer to make that high-watermark payment each subsequent year.

Second, there is no inflation adjustment during these periods (in this case seven years plus) when the income is frozen. This represents a diminishing income stream.

And third, for most of these annuities, the yearly cost of the rider is higher than you might think. In many cases (including all Nationwide annuities), if the portfolio drops, the fee will be calculated as a percentage of the high-watermark value that was used to calculate the distributions. So after the bear market has done its damage, the study (correctly) calculates the 2003 cost of the GMWB rider as 60 basis points multiplied by $3,831,558-the portfolio's 1999 value. If you figured this as a percentage of the $2,233,264 that is actually in the VA portfolio, the cost amounts to about 1%.

I began to suspect this research was slanted in favor of VA contracts when I saw Table 2D, which offers some assumptions behind the spreadsheet and various flattering comparisons between the VA with a GMWB rider and mutual funds. Funds and VAs are assessed a 1% yearly drag for portfolio management expenses and are docked 1% a year for a planner's services.

Unreal Assumptions?

Then comes the part that had me scratching my head. The study assumes a 0.4% mortality and expense (M&E) fee each year, and a GMWB rider fee of 0.6%. Come again? Morningstar lists an average M&E charge of 125 basis points (bp); some of the most popular VAs charge more. To take some not-so-random examples, Nationwide's Elite Venue Annuity has a 1.55% M&E charge, plus an annual 20 basis-point administrative fee—for a total of 1.75% a year. Best of America All-American Gold Annuity's M&E is 95 basis points, plus the 20 bp administrative fee. Nationwide's Future Venue annuity: 90 bp M&E, plus 20 bp administrative expenses. Best of America V: 1.1% M&E. Best of America Achiever: 1.35% M&E, plus 0.2% administrative charge.