Friday, November 23, 2012

Important New Research Results from IFID Centre Conference

I'm writing from Toronto, where I attended the 2012 Annual IFID Centre Conference orgazinized by Moshe Milevsky and his colleagues. There were four speakers at the conference, each of whom provided research results which have a direct bearing on retirement income strategies. The real mind blower for me was the fourth and final presentation, so don't click away yet.

I was the first speaker. I spoke on my most recent research article, "An Efficient Frontier for Retirement Income." As this is a research area pioneered by Moshe Milevsky in the first place, I was nervous about doing this. But I think things went well. I've discussed this paper here before. At the conference I also showed results from further case studies which I had been meaning to write about at my blog but haven't gotten around to yet. These include looking at different measures of success for meeting lifestyle spending goals, assuming a smaller equity risk premium of 3% (which Robert Kirchner had encouraged me to do), and also trying a more optimistic scenario for stock and bond returns to make sure that my previous findings are not only a result of my pessimistic assumptions about market returns. In all the cases, the general shape of my efficient frontier holds up, which is that retirees may be best served by combinations of stocks and fixed SPIAs. A part of the frontier I hadn't discussed also does include that particularly risk averse retirees may benefit instead from combinations of real SPIAs and fixed SPIAs. These results do appear relatively robust. The big takeaway for me from the participant comments is that I should try including life insurance as another potential "asset class" for the analysis. 

The second speaker was Anthony Webb from the Center for Retirement Research at Boston College, who spoke on his research article, "Should Households Base Asset Decumulation Strategies on Required Minimum Distribution Tables?" I recently discussed research from Morningstar about optimal withdrawal strategies, and their optimal strategy is rather complex. Webb and co-author Wei Sun limit their analysis to "rule of thumb" style strategies which can be more easily followed by households, such as the 4% rule, and other related strategies. They find relatively robust results that a good "rule of thumb" strategy is to spend any interest and dividends received plus a percentage of principal equal to the amount calculated by the IRS for the required minimum distributions from tax-deferred accounts. I discuss those RMD distributions and provide a table of them here. My only concern about this strategy is that it provides a variable and unpredictable income which is highly dependent on market returns, but Anthony Webb's intuitive response is that: if you don't want to accept variable income, you shouldn't be investing in the stock market in the first place. That's a safety-first idea

The third speaker was Marie-Eve Lachance from San Diego State University. She focused on how to optimize between traditional IRA (backloaded taxes) and Roth IRA (frontloaded taxes) accounts in light of unknown future tax policies and the complications arising with other parts of the tax code, in particular the fact that traditional IRA distributions count as income for whether Social Security is taxed, while Roth IRA deductions (currently) do not. Conventional wisdom is that people should contribute to tax deferred savings accounts first with the idea that taxes will be lower post retirement. But this result does not always hold up, and college educated readers (implying a higher salary level) may benefit more from frontloaded tax approaches. Here is a link to her rather theoretical article on this subject.

Finally, Thomas Salisbury of York University spoke on work-in-progress research by him, Moshe Milevsky, and Huaxiong Huang. This is the presentation which I found myself most fascinated by and took copious notes. Though it is still a work in progress, I received permission to write about it. The maestro himself, Moshe Milevsky, will also write a column about it early next year for Research Magazine, and the authors, of course, will be sending their research paper to an academic journal as well. So definitely watch for that (I will surely provide a link when it comes out), but you are hearing about it here first :)

What is the issue? Well, it is a rather specialized topic and may not be relevant for many readers. But for those of you for whom it matters, this is a big deal. And I know some of you care, because I've received emails with questions about this issue and I didn't know a good answer. If you would like a general introduction to what I am talking about here, I provided an overview of GLWBs in a recent column at Advisor Perspectives.

The question is: suppose you own a deferred variable annuity with a GLWB rider and you have not begun taking guaranteed withdrawals yet. You are still in the deferral phase, but you are specifically wondering when you should stop deferring and start taking guaranteed income. These co-authors worked through the mathematics about this and found that the general answer is simple: start taking guaranteed income ASAP. There are some exceptions, but most of the exceptions are not very realistic for being actual possible scenarios for people. The one exception which may matter is that if you are within a couple of years from reaching an age where the guaranteed withdrawal rate will jump up (such as from 4% to 4.5%) then it could be worthwhile to wait for that. Their research paper will provide the exact parameters about when this wait is worthwhile. But otherwise, generally speaking, if you are thinking about whether you should start taking guaranteed income, the answer is going to be that you should start taking guaranteed income now

They investigate this question from the perspective of what is the worst thing a GLWB owner can do from the perspective of the insurance company. In turns out that especially in the past, many GLWB products seem to have been designed to be lapse-supported. This means that the GLWBs actually offered pretty valuable guarantees, but the saving grace for the issuer is that most of the owners would let the product lapse or would roll over to a different annuity before the insurance provider ever found themselves on the hook for supporting guarantees. This all changed with the financial crisis in 2008, after which a number of companies have dropped their GLWB lines as owners found out that they now owned valuable guarantees that were in the money (the guaranteed benefit base exceeds the contract value of remaining assets) and were not lapsing at the rate the insurance companies expected. Compounding the trouble for the insurance companies was that with interest rates falling so low, it was becoming increasingly difficult for them to hedge the risks related to the guarantees they had made. 

And so suppose you own a GLWB in the deferral phase and are thinking about starting income. The intuition for why you should start income now is rather elegant. GLWBs are complicated, but they can be thought of simply as a combination of an investment account and a deferred annuity. As long as the assets in the VA account are bigger than zero, then withdrawals do not cost the insurance company anything. You are taking back your own funds. The GLWB becomes costly to the insurer (which is good for you) in the case that the contract value of your assets reaches zero. Then you still receive an income for life and this is the expense which the insurance company needs to be hedging against with the income they receive from the guarantee rider they charge you. So from the insurance company's perspective, they want you to delay. And they provide a lot of incentives to do this, such as the opportunity for growing your benefit base if your investments do well, the opportunity to get a higher payout rate if you wait to start at a higher age, or even some companies provide guaranteed growth for your benefit base if you delay. 

But the opportunity to grow your benefit base by delaying guaranteed income just isn't worth it (except for that issue where you are just a couple of years away from getting an age-based increase in your payout rate). Think of it this way. Suppose you delay taking $100 for one year and this lets your benefit base grow by enough so that your subsequent guaranteed income will be $105. Assuming that the interest rate for discounting future income is zero, how long do you have to wait for this deferral to pay off? Well, it doesn't pay off until the contract value of assets hits zero. Suppose that takes 25 years to happen. Then the deferred annuity begins which is $5 more per year than otherwise. Then, it takes another 20 years to get that $100 of income back that you had delayed. You are now looking at a 45 year period to breakeven on your delay. With a positive interest rate, the breakeven point would be even longer. It just isn't worth it. 

I think this result is so powerful because it is not intuitive beforehand, but makes a lot of sense once you hear it. Do note that this issue is different from whether or not you should buy a GLWB in the first place. The matter is what to do if you happen to already own one. And the answer is: start income withdrawals!

Attending the conference was a great experience. It is held annually on US Thanksgiving Day.


  1. Any thoughts on the RMD approach using a rolling 3 or 5 year average year end balance to smooth fluctuating spending pattern?


    1. Yes, that approach could make sense to smooth spending fluctuations.

      This is an example of a type of additional complication which Sun and Webb did not consider, since it wouldn't be very easier for a typical household to calculate on their own.

      Another approach would be to modify the RMD tables to assume some percentage growth of assets. The RMD tables assume that assets don't grow, which makes them overly conservative as a withdrawal strategy for most people at the start of retirement.

  2. as always, really helpful information. Thanks Wade. Looking forward to your new efficient frontier case studies. Updegrave suggests having 25-30% of bonds in TIPS in case inflation spikes (noting that regular bonds do better if inflation is mild). I wonder if the same strategy may be a helpful hedge with SPIAs : having 25-30% in real SPIAs, 70-75% fixed : just in case some time in the next 30 years there is higher than expected inflation (or dread: hyperinflation).


    1. Thanks. This is a good point.

      Current market expectations are that inflation will stay rather low in the future, but to the extent that you think this is too optimistic, then it is certainly worthwhile to include more inflation protection components.

  3. The work on GLWB seems to imply that there are no relevant changes at the time one begins receiving income. I can think of at least three possible problems with these assumptions, although I don't know how often any of them would apply: (1) Some contracts have guarantees on rate of return in the accumulation stage that might play into this decision. (2) Do any contracts have different rules for investment choices between the accumulation stage and the payout stage? (3) Are there differences between "step up" rules in the payout stage and changes in balance in the accumulation stage? Note that those holding for another year might find themselves with lower balances a year later! There may also be other considerations. Can you think of any? I guess I'm just concerned that such simple advice may not be appropriate for all the varieties of such a complex product.


    1. Thanks for the comment. About your suggestions

      (1) They have accounted for this and find that realistic levels of guaranteed returns for accumulation are not sufficient to stop the decision to delay

      (2) That's interesting. I'm not an expert on the various versions, but I haven't come across any products that had different investment rules for accumulation and distribution. I don't think they accounted for such differences. Do you mean that distribution is more conservative?

      (3) I don't think that will matter. For their math, they assume that step ups occur instantaneously when the guaranteed base rises, but I don't think it will make much difference for quarterly, annually, etc. There are not going to be many step ups once distribution begins anyway.

      About your point that the account balance declines. That is true. But you don't have to spend the distributions. You could "save" them again by moving them to another investment account. The point is, once you have a GLWB, the only way you benefit is by withdrawing to the max to get the account balance down to zero as quickly as possible.

      One issue that they didn't account for is taxes. But other than that, I don't know of any other considerations.


  4. Wade,

    It makes me giddy with joy that, once again, you took the time to document original retirement research findings for public/mass consumption. As someone whose job includes thinking about retirement planning - and GLWBs in particular - there is no substitute for the prolific work you are doing here. It's not just your original research, but your willingness (eagerness?) to share others' ideas. There is a non-zero probability I would not have seen many of these articles otherwise. In my limited time I have to devote to reading the literature, this blog makes an amazing one-stop destination. A thousand times, thanks.

    1. Thanks Daniel,

      I appreciate it! Aiming to make this into a one-stop destination to learn about the latest research on retirement income sounds like a pretty laudable goal. I'll work on that!


  5. Wade - Thank you for this info! My question is, you state "The point is, once you have a GLWB, the only way you benefit is by withdrawing to the max to get the account balance down to zero as quickly as possible."

    I have a GLWB and I was advised to take it to maintain my benefit base and if my account value ever gets so low that it will not support another withdrawl to then annuitize the benefit base. You make it sound like I should take distributions greater than what would be had to maintain the benefit base which in my case resets up by 6% on every contract anniversary date. So, I should request the maximum allowed? Thanks!

    1. Hi, I am not sure I fully understand the wording of your question.

      You do not want to withdraw more than what is allowed to maintain your guaranteed income. But the findings from Moshe Milevsky's new research is that yes, you should withdraw the maximum allowed. Even having the benefit base otherwise grow by 6% a year is not enough to make it worthwhile not to withdraw the maximum allowed income.

      Note that this is new research. The idea that having a benefit base grow at 6% does sound nice. So the person could have honestly recommended that to you. But it turns out that it just isn't worth delaying income for that reason.

      Still, before taking this advice, be sure to discuss it carefully with your advisor. These are research results that may not apply to everyone's situation.