Tuesday, September 24, 2013

Why Retirees Should Choose DIAs over SPIAs

My new column at Advisor Perspectives is called, "Why Retirees Should Choose DIAs over SPIAs." In the past, I've written about the efficient frontier of retirement income, finding that retirees can best satisfy twin goals of preserving their lifestyle spending needs in bad luck cases and leaving the potential for upside as well through a combination of stocks and SPIAs.  In this new column, I address the issue of what happens if deferred immediate annuities (also sometimes called deferred income annuities or longevity insurance) are added to the universe of choices when calculating the efficient frontier.

DIAs are like single-premium immediate annuities (SPIAs) in the sense that a lump-sum payment is made now to the annuity provider in return for a guaranteed income stream for the remainder of one's life. The difference is that with a DIA, that guaranteed income stream begins at some later date further off into the future, rather than within one year. This allows the DIA to provide longevity protection more cheaply.

In the column, I describe what happens when I add DIAs with different deferral dates into the mix of retiree choices. I find that DIAs with 10-20 deferral periods [i.e. a 65 year-old couple buys a DIA with income beginning sometime between age 75 and 85] result in more efficient outcomes (better downside protection with very little impact on upside potential) than SPIAs. As well, including SPIAs or DIAs result in noticeably better outcomes than leaving everything just in stocks or bonds. Please see the column for more details.


  1. Hi Wade
    An interesting piece of analysis in the paper. It strikes me that there will be a big difference in growth allocations across the total portfolio. I am wondering (much like your recent paper with Michael Kitces) if this growth allocation is impact the results. Having 68% in the SPIA leaves only 16% in growth assets which will hamper the median growth overall. The 20 year defered DIA will still have 41% in equities.
    How would it look if you set total exposure to 30% growth at the start across the scenarios and then maintained that allocation

    1. Aaron,

      That's an interesting point and I should re-run the simulations with that in mind. I do know the 50/50 portfolio for financial assets is not optimal, but as the optimal financial portfolio tends to be 100% stocks, I know that isn't particularly realistic and so just tried to use something more realistic.

      The implications of what you are suggesting though, is that I'm penalizing shorter deferral periods because they will have a smaller absolute allocation to stocks. It's interesting to check.

  2. To get around many of the inherent problems with DIAs, I'm building my own DIA-equivalent instead and encourage you to consider this strategy in your research.

    I'm doing this by (a) purchasing long-term zero coupon treasury bonds (STRIPS) over time (dollar cost averaging) that mature when I turn 85 (my wife would be 80), (b) planning to buy a joint-lives SPIA with most or all of the funds when I turn 85. Currently, I'm 27 years away from that age and the YTM for the zeros being purchased now is about 4%. The purchases -- inside a Roth IRA -- will be largely completed within a year or two (when long bond rates perhaps more fully reflect a winding down of QE3).

    Some advantages of this strategy over buying a traditional DIA decades in advance of age 85: (a) I retain full access to the bonds until the SPIA is purchased, (b) we will have much more information about life expectancy for each of us when the annuity is purchased (and can thus choose a single-life SPIA, skip the SPIA, or seek an impaired life SPIA -- if one or both of us has health conditions that make it unlikely to outlive the predicted remaining years in the mortality table), (c) we can optionally skip, scale back, or even choose a slightly earlier start for the SPIA depending on the condition of our household balance sheet at age 85 and awareness of expenses that could not be known accurately decades earlier, (d) if nothing else, the dollar cost averaging helps to avert buyer’s remorse, and (e) we are not depending on the solvency of an insurance company for decades before the payouts start; we are depending on the solvency of the US Government, but so are insurance companies.

    The IRR for the homemade DIA isn't quite as robust as a commercial DIA purchased many years in advance (about 75 basis points less per year from purchase), but the gains in flexibility from the homemade DIA make up for that.

    1. Thank you for sharing. Your strategy sounds reasonable to me. I would like to study more about the possibility of waiting until a later date to purchase a SPIA, but a fundamental problem is not knowing what interest rates (and, therefore, SPIA rates) will be in the future. Most existing research which looks at this projects today's interest rates forward. I suppose, in some sense, this unknown future interest rate risk with waiting to buy a SPIA would be counterbalanced by the unknown future inflation risk associated with buying a DIA today. At any rate, I don't see any obvious problems with your approach. Thanks.

    2. Hi again, Wade--

      In my example at age 85 above, the mortality tables show only 12 years of joint life expectancy remaining, so most of what the SPIA would be paying out is a return of premium.

      Put into numbers, my SPIA would pay out $89,679/yr in 2040 with IRR=3%, $94,679 with IRR=4%, and $99,733 with IRR=5%. In today's dollars, with a 3% discounting rate for inflation, those numbers would be $40,372, $42,623, and $44,898 respectively. So the real annual payout difference across even a 200 basis point change in IRR is rather modest.

      I find that people often get dazzled by the apparently high nominal payouts of DIAs decades from now and that advisors and insurance companies are reluctant to spoil the fun by translating those into present value dollars.

    3. Thanks. That's an interesting point about future interest rates not making a big difference about future SPIA purchases, since mortality credits and return of principal become so predominant. Actually, that is a really good point. Future interest rate assumptions don't matter so much. Maybe I can add this into my investigations after all.

      I agree with you about inflation. Even modest inflation, when compounded over a long enough period of time, can make a substantial difference.