Thursday, October 10, 2013

Comparing SPIAs with Managed Accounts

Larry Frank, John Mitchell, and I have finished a research article called, "Lifetime Expected Income Breakeven Comparison between SPIAs and Managed Portfolios." This post is the third is a new weekly tradition of a video blog post made in cooperation with The Wealth Channel at the American College. I summarize the article in the video:

For email readers, the videos never show up in the email, but you can see the video by clicking here.


  1. I cannot wait to completely read your new research article, but I have a concern/question about the ‘Overall Discussion’ statement on page 4:

    Fundamentally, a retiree has two basic choices about how to fund their retirement income; 1) to annuitize their retirement assets, or 2) to manage those assets, either by themselves, or through assistance of an adviser.

    What about the third option of having a mix of SPIA and non-SPIA investments?

    Now, as an individual investor, I don’t like SPIAs as a retirement investment, but I do see the desire/interest by others. They are not right for me and my family’s retirement plan.

    I thought the common rule was to have no more than 25% of retirement investments in SPIAs. I also thought a common plan of purchasing SPIAs in increments was done to counteract the issue with varying SPIA interest rates. For example, one could begin purchasing a SPIA with 5% - 10% of their retirement assets at age 60-70 and then purchase other SPIAs every five years at increments of 5% of total assets. I agree that age 85 is the ideal age to own SPIAs, if appropriate (i.e. longer expected longevity).

    Purchasing SPIAs in increments would also allow them to spread the risk of having the investment company falter; not that any companies have ever faltered, but it could happen and as I understand SPIAs, they are not protected by government regulations like our normal retirement investments and savings accounts. Having five different companies take all of your SPIA investments is highly unlikely.

    Again, I do not recommend SPIAs, but if someone does want SPIAs I think they should consider a tapered approach with only a small portion of their nest egg.

    Yet another great article. Keep 'em coming.

  2. Paul,

    Thanks for the feedback. I agree that this article presents the decision as either-or, when in reality a more balanced approach is is more advisable. I'm still hoping to look more at the possibility of laddering / delaying SPIA or DIA purchases as part of an overall strategy. Thanks again.

  3. I usually find myself basically in agreement with your papers. But I have some issues with this one. My concerns are not with the calculations – I haven’t checked them but they came out about as I would expect. After all, insurers buy bonds to hedge their liabilities for annuities; given insurer overhead and, usually, profit, one would expect an annuity on the average to be less favorable than a direct investment in bonds and well behind the expected value of a portfolio containing equities. But the first bullet point says “purchasing a SPIA at any age is unadvisable” Doesn’t that contradict your earlier research on the efficient frontier? In addition, if I understand the article correctly, it biases the data against SPIAs by using historical returns for other asset classes but current (i.e., low) returns for annuities. (Unless, of course, you think that future stock and bond returns will on the average match past returns. Haven’t you been a skeptic about that in the past?) In addition, comparing expected values or 25th to 75th percentiles of the cash flows seems odd. Annuities are insurance products. Do you recommend that people not insure their homes because the expected value of the insurance is negative and because over 95% of those purchasing insurance will lose money on the transaction? The value of an insurance product is in the unfavorable tail of the distribution, not in the middle. Also, you write “[T]he insurance company may also be exposed to the risk of going out of business trying to sustain high payments in a poor market environment that those higher relative payments were not priced on.” and you express similar concerns several other places. This language makes it sound as if the insurance company takes the money from the annuity, invests it for few years, and then hopes to reinvest at the same or higher rates later. But, at least for an SPIA, I believe that in fact the company hedges the expected lifetime risk immediately by acquiring bonds to match the entire lifetime stream of liabilities. That seems to be confirmed by Charupat, Kamstra, and Milevsky in , where they observe that the actual duration of annuities is close to the theoretical duration. Of course, the companies can’t match perfectly because future mortality is not known, so there may be some minor adjustments later; but they should be indeed be minor. I know several actuaries have commented on this blog; perhaps they can tell me if I’m wrong about that? In general, this article reads like “Let’s make the strongest possible case against annuities” rather than as an objective evaluation.


    1. Thank you for the feedback.

      You are right about “purchasing a SPIA at any age is unadvisable” being misplaced. That is not the finding of the article. We've been revising it a lot since posting the SSRN version and that has been removed.

      You are also right about the unfair comparison from using historical averages for other asset classes. I've argued against doing that in the past. In this case, my co-authors feel this it is justifiable to use historical averages, though I tend to disagree on this point. Especially when it comes to comparing the historical average bond yield to the current bond yield.

      I also agree with you about the point that the 25th percentile is not sufficient.

      Essentially, this article does move in a useful direction by trying to compare annuities to systematic withdrawal strategies which make adjustments in response to market returns, but I would agree with you that more needs to be done with the framework. That is why, in the video, I focused on the cases when annuities do well, despite the handicaps, than when the paper makes it seem that systematic withdrawals could seemingly do better. Admittedly, I could have spent more time fleshing out that point and making it more clear, though.

      Thanks for the feedback and for making these important points!

  4. Statistically speaking, perhaps you are correct that purchasing SPIAs at any age is unadvisable from an expected value of lifetime cash flow perspective. However, this ignores the fact that people largely purchase SPIAs – and other insurance products – for peace of mind, not necessarily to maximize total financial return. Many elderly people, even with substantial portfolios, would not experience much peace of mind after a sudden large drop in portfolio value due to bad markets. And it would be little consolation at that point for them to ponder how historical market returns – based on Monte Carlo simulations – are in their favor. They may never live long enough to recover from a large market loss and this could be psychologically devastating even if they have plenty left to cover expenses. If a key retirement objective is peace of mind – rather than maximizing cash flows – then purchasing SPIAs can be very rational indeed.

    It is also noteworthy that there is no mention of SPIAs based on *joint* lives (typically covering an annuitant + spouse, with most or all of the payments continuing to the survivor) and the effect this may have on your calculations. For instance, after one spouse passes, it is possible that the surviving spouse on a SPIA may have more cash flow than necessary due to reduced living expenses. This could be accumulated for bequest motives. And contrary to Reichling and Smetters’ (2013) assertion that health shocks reduce the PV of the remaining annuity payments in a SPIA, this may not be the case at all if the SPIA is based on joint lives and the health shock only involves one party.


    1. Thank you, you are making good points. I especially appreciate the points about how matters may differ for joint-life annuities. I haven't seen those points expressed before!

  5. As we discussed at the Bogleheads conference, and as mentioned in comments above, it is obviously more difficult to model, but it seems a better approach would be to annuitize a small percentage of the portfolio every few years and then to adjust withdrawals from the remainder of the portfolio according to market conditions- i.e. take out less in 2008-2009 and take out more in 2012-2013. A common-sense approach such as that would seem to make sense for the vast majority of individuals, even if it is difficult to study.

    1. Thanks. I think it will be practical to model this if we can simply assume that interest rates in the future will be the same as today. Otherwise it gets too difficult to know what future annuity payout rates should be without having a model that can successfully simulate both bond yields and bond returns at the same time.