Wednesday, March 7, 2012

To annuitize or not to annuitize

Just a short message here...

Joseph Tomlinson is on a roll. His new article at Advisor Perspectives called "New Tools to Manage Longevity Risk" provides an exploration of deferred income annuities (which means, pay now but defer the date that benefits arrive). It's an interesting article related to what I recently discussed in "Safe Retirement income with TIPS and a deferred annuity." I also appreciate more and more how his Monte Carlo simulations start with inputs for the current bond yields (not historical averages) and the average guess of many "experts" about what the future equity premium will be.  This does result in lower success rates for the 4% rule than just basing the simulations on historical averages. Especially considering that I wrote "Can We Predict Sustainable Withdrawal Rates for New Retirees," I'm guilty of basing Monte Carlo simulations too frequently on historical averages as a convenient shortcut. Joe is doing it the right way.

Second, in the past year I've had an opportunity to develop some valued friendships with individuals I've connected with through research. Bob Seawright writes frequently about the value of single-premium immediate annuities for one's portfolio, and my research does tend to validate this. I'm still researching annuities myself, but I am coming to think that partially annuitizing one's nest egg may make sense for a lot of people. Another friendship I've developed is with Dan Moisand, who recently cited my research to help make a case against annuities in his article "An Unattractive Proposition" at Financial Advisor magazine. The research he cites is a blog post "Retirement Withdrawals and Leftover Wealth" in which I describe about an issue Bill Bengen mentioned in an interview with Forbes last May.  


  1. I really appreciate the contacts I've made with Wade highlighting my work, but what I appreciate even more is that, in this blog, participants can feel free to say things like, "I'm not sure I'm looking at this correctly." In that vein, I'd like to note that I'm in the process of trying to sort out some confusion in my mind about bond returns and the equity premium. There was a good article in Advisor Perspectives a month ago that I missed. There's some confusion over the equity premium because predictors often don't say whether they are talking in geometric or arithmetic terms and there's about a 150 basis point difference (3.5% geometric is about 5% arithmetic--I think). Also, on real returns, I get confused myself about whether to use current real returns (close to zero) or what I expect over the long term (maybe 2%?). Perhaps, if I'm evaluating an investment today, I use zero, but if I'm looking at long term financial projections for a client, I should assume that real rates won't stay at zero forever, and will increase over time. (If I were more adept, I might use the current yield curve to predict the future course of rates.)

    Where I've been a bit confused is I thought I was predicting a lower equity premium going forward. What, it seems, I have actually been doing is making more pessimistic predictions because I been using current real interest rates.

    For this latest Advisor Perspectives article I used a zero real interest rate assumption and a 5.5% (arithmetic) equity premium. Based on the Laurence Siegel article, I'm wondering if I should drop the 5.5% to 4.5%? I'm thinking of holding to the zero % real interest rate because I'm doing comparisions to annuity and DIA products that reflect current rates. Any thoughts would be appreciated.

  2. Most of the work on annuities I have seen has concentrated on either immediate nominal annuities, immediate inflation-adjusted annuities, or annuities with guaranteed minimum withdrawal rates. But there is another reasonably common form of annuity that is tied to the performance of an underlying investment but does not provide a guarantee and allows more investment choice and has lower costs than ones that do have guarantees. The canonical example is CREF; another example is the graded TIAA method, although that appears basically defunct for now because of low interest rates. A more generally available example with similar costs is the AIG annuity available through Vanguard that currently has a mortality charge of 52 basis points; of course, the "all in" cost would include the cost of the underlying mutual fund(s), often quite low (20 basis points or so). There are many possible underlying funds, including full equity (generally not available with a guaranteed withdrawal) and a high-yield bond fund. Has there been any research on the use of this kind of annuity?

  3. I think the Vanguard product is now American General. It's a variable immediate annuity. The idea is it pays an income for life like the more popular fixed immediate annuity, but the payments vary from year to year based on investment portfolio performance compared to what is called an Assumed Investment Rate (AIR) chosen by the purchaser. So payments can go up or down. It's not ideal if one is using the annuity to pay for basic living expenses, but if income bouncing around is OK, then it can do better than a fixed immediate annuity. In its most basic form, it provides no liquidity or death benefit, but that means it can provide a higher average level of income than the standard variable annuity with guaranteed lifetime withdrawal benefit. (The level of fees is important too in any annuity comparison.)

    I don't know about sales levels for TIAA-CREF, but in general the product doesn't sell that well--probably too complex for most people. As for research, I'm aware of a study Olivia Mithell was involved in:

    It's pretty heavy on math, and I should probably sit down and read it again one of these days. It's been a few years since I read it. I'm not aware of any easier-to-read studies, but if you Google variable immediate annuity, you might come up with examples. It's worth studying if one wants to compare all annuity alternatives.

  4. Joe and Palc,

    Thanks. Palc, as I'm still in the learning phase about annuities, I appreciate Joe reponding to your question.

    Joe, that Advisor Perspectives column is very interesting and useful, thanks.

    About your concerns:

    I do think that using the current yield is most appropriate: both because the process of returning to a higher "normal" yield would involve capital losses, and because due to the sequence of returns risk in retirement, what happens earlier on in retirement weighs disproportionately on the final outcome.

    But if those equity risk premium estimates are calibrated to a higher average yield instead of the current low yield, then that could be a problem. Also, whether they are talking about arithmetic or compounded equity premiums is also important. I interpreted what you were writing as discussing the arithmetic equity premium. I had the same problem with Van Harlow's article, as I never knew for sure if he was describing an arithmetic or geometric stock return, but assumed he meant arithmetic.

    For the Monte Carlo simulation program I wrote, I input arithmetic returns and standard deviations, and my program takes care of converting that into geometric returns.

    So in conclusion, I do agree that you are making a very important point that when people make equity risk premium estimates, they need to clearly state whether they are comparing to current or average bond yields and whether they are describing the arithmetic or geometric premium.

  5. Why use economists guessed when the guesses are not predictive of much of anything historically? Which is why I assume you put the "experts" in quotes to begin with?

    "average guess of many "experts" about what the future equity premium will be"

    When I read the financial history, it just seems that humans, economist or otherwise, cannot predict much of anything that will happen financially.

    Thanks for the blog, good reading,


  6. LH, thanks. Yes, that is why I put "experts" in quotes. It is hard to be an expert on the subject of predicting the future.

    But I do understand the need that some sort of average assumption is needed to explore more about investing strategies with Monte Carlo simulations. The average guess may be as good as anything in this case.

      annuity rates,
      best annuity rates,
      annuity quotes,
      fixed annuities,
      bonus annuities