Thursday, February 14, 2013

Larry Frank Guest Post on Viability of Annuity Providers


This is a special guest post of an email exchange I had with Larry Frank, a retired Air Force Command Pilot and Contingency War Planner, and following that, almost two decades now in the financial planning profession as a Registered Investment Adviser and owner of his practice Better Financial Education in Roseville, CA. He blogs about retirement and many other financial planning topics at Blog.BetterFinancialEducation.com.

He focuses on how to develop plans for prudent and sustainable retirement income for the average middle class person to get them both to, and through, retirement. This starts with helping families determine how to maintain their standard of individual living which is specific and unique only to the them, i.e., no generic income replacement ratios.

Larry is the author of Wealth Odyssey and 5 peer reviewed research papers in the Journal of Financial Planning. He holds a BS cum laude in Physics from the University of Minnesota (how things work) and an MBA from the University of South Dakota (how money works). He is a CERTIFIED FINANCIAL PLANNERTM Practioner.

A list of his published research articles can be found at his website

Larry wrote to me with comments about my efficient frontier article in the February Journal of Financial Planning. He makes an important point that in its current form, the efficient frontier is still static rather than dynamic as I have not built in ways to adjust away from the fixed lifestyle goal. Larry's research has focused on building more dynamic strategies for safe withdrawal rates which adjust for both age and portfolio experience. I haven't read his series of articles in the journal yet, but I did review a complementary version of the approach by his co-author David Blanchett with colleagues at Morningstar. 

Larry describes himself as part of the probability-based school of thought rather than the safety-first school of thought. As we discussed it more, he wrote some rather provocative and thought-provoking comments about the role of annuities which I haven't seen expressed in such a way before. I thought his comments were interesting enough to warrant a guest post. In an earlier message he expressed concerns that insurance companies were not properly incorporating longevity improvements into their calculations (i.e. people will live longer than insurance companies expect, which could jeopardize annuity guarantees). I asked him what he thought about the idea that insurance companies could hedge this longevity risk by selling both annuities and life insurance. If people live longer than expected, annuity payments increase, but life insurance payouts decrease.  Here is his response:
It seems to me Wade, that hedging is a short term strategy (5 years at best, with the majority only months) while the obligation to pay a benefit may last decades for an annuity. At the moment, immediate annuities are such a small percentage of assets that any blow ups can be almost absorbed by the insurance company. However, with the great push by legislatures and the insurance industry to “solve the retirement problem, ”the volume of assets may get so large that it becomes too big to fail.  The promised payouts may exceed assets on such a large scale that it overwhelms both insurance companies and State guarantee associations.  The solution doesn't match the problem here, where the real problem is people not saving enough in the first place, yet the “solution” doesn’t address that! Instead, the “solution” is to put more immediate annuities into 401ks, thus leading to a future increase in assets.
To big to fail? How would this happen? By the tendency to promise more out of the system than the system can sustain. Our last paper (JFP Dec 2012) opened my eyes to the fact that if the amount of assets shrinks, then the sustainable amount distributed has to shrink also (and vise versa). You can’t get more out than what the assets are capable of supporting. The tendency is to get too generous during good years (look at what happened to variable annuity benefits riders) and then it gets hard to take those back. Better to provide bonuses during good years rather then set up an income for life based on good years and ignoring the possibility for lean years.
Increased longevity should reduce annuity payouts because a set sum needs to support more years. So the payouts are based on longevity today, but while that retiree ages the longevity extends … thus, more is promised to be paid out than can be sustained. Ramp this up by many more retirees promised beneifts through their 401k's, and the total assets can’t sustain a longer living group. This effect combines with the above promises effect, and the system is not sustainable. The insurance industry made faulty assumptions about the VUL product in the late 90’s and faulty assumptions about VA riders in the 2000’s. Their track record based on “old” concepts suggests an insidious set up could take place … similar to the slow insidious forces that set up the mortgage industry … unless assumptions and models aren’t critically looked at.
When I talked with a number of professors at the Academy of Financial Services who are researching retirement, they say getting good data about current strategy, etc, for academic research is difficult because insurance companies tend to think of it as proprietary, or don’t want the academic scrutiny. So getting a good broad academic look at the topic is challenging.
A dynamic approach that adjusts payouts, like the dynamic model in our paper series, based on current asset levels provides an asset level that can be sustained into super centenarian ages. Income cannot be outlived if the supporting assets are properly managed as to sustainable payouts. However, the goal seems to be to promise a guarantee, which often ignores asset reality, and we could easily get into problems that pensions now have.

37 comments:

  1. Interesting discussion. Could one of the actuarial problems be that actuaries are aftcasting rather than forecasting? If people are living longer than expected, then something is wrong with the set of assumptions which led to the expectation in the first place - a happy result for the survivors, but an unhappy result for the actuaries and insurance companies. I'm not an actuary, so I don't know how they work, but if they're using past performance as a guide, then, because of advances in medicine, healthcare, etc., aren't they nearly automatically going to undershoot the expectations of living? The implications would mean that current generations of buyers would be funding past generations of still kicking buyers and the insurance company would have to hope for outsized market performance in their investments to cover the gap. Of course, if actuarial tables are built with forecasting in mind, then this entire hypothesis is moot.

    ReplyDelete
  2. Sort of a side question, I guess, but in terms of annuity safety: I get the need to diversify across issuers if one is concerned about issuer-risk, but how safe are SPIA's if the underlying asset-classes have major periods of underperformance? Given the major mistakes that big insurance companies have made in recent years in their projections on diasbility and LTC insurance, I am not necessarily confident in their actuarial prowess.

    ReplyDelete
  3. The actuaries try to do as good a job as they can … prediction is impossible, so by nature the math is historical (I’m not an actuary either by the way). There are actually two sets of probability that need to be combined (JFP March 2012); one, probability of the portfolio (sequence risk), and two, probability of the person (longevity risk). However, the topic here of mortality is a matter of perspective as to what is happening slowly but surely over time. As I mentioned in my exchange with Wade, the present mortality tables are based on people born, for example in 1933 (or even earlier), dying now to populate the tables with death rates. When they retired say in the 1980’s, that mortality table then was different than it is now. Because those who died to establish the mortality table in the 1980’s had a different experience with their health and medicine than we do today. So those retirees experienced a changing mortality table already. The mortality table has extended expected longevity for those retirees. The same forces are in place to continue to slowly extend the table for those baby boomers, born in 1953 for example, retiring today. Because expected longevity is extending, distribution periods are slowly extending. The 20 year difference between birth years is resulting in more than a 20 year difference in expected longevity. And that extension may continue, so those born in 1973 for example, would have an even greater extension to expected longevity than the previous 20 years. We won’t know what expected longevity will be 20 years from now, so trying to forecast it is not possible. However, by dynamically updating the distribution periods based on current period life tables, as they are updated over the coming years, will compensate for this unknown slowly over time. An older expected longevity extends the distribution period, which means that withdrawal rates need to be revised down slightly over time to so the balance may last over a longer period. Annually recalculating both the expected longevity age and using current portfolio balances that incorporate what market returns (gains or losses for the year) have done, keeps the income distributed connected to the asset based designed to support that income. This is what I mean by the term dynamic. Our paper in JFP (Dec 2012) showed how this can be done even into centenarian or super centenarian ages. At present, promised income benefits from pensions, annuities, and implicitly in the first generation application of the safe withdrawal rate concept disconnect the income promised from the benefit base supporting those benefits. If the volume of those assets becomes systemic, then the scope of disconnected income promises may become a problem (too big to fail?).

    ReplyDelete
  4. I am not an actuary either, but this difference is something that we see with all the time working with advisers and retirement income plans. I usually refer to the static historical basis for estimate life expectancy as the statutory basis. The reason is that this basis is used for tax and social security calculations. For a 65-year-old Australian male, life expectancy on this statutory basis (2005-07ALT)is 84.
    Adjusting for improvements in mortality is done with a trend improvement in mortality. This adjusted life expectancy for a 65-year-old Australian male is currently 87.
    The actuaries in life companies (at least in Australia) do know these numbers, and also adjust for adverse selection (unless annuitisation is compulsory, people with terminal illnesses, for example, will not buy a lifetime annuity). The payouts are based on expected longevity.
    Having the assets to support the future income stream is important for life companies as well as any individual retirement income stream.

    ReplyDelete
  5. What Aaron says about Australia also applies to the US. Actuaries develop explicit mortality improvement assumptions that they use in pricing. They are not developing mortality assumptions by looking in the rear view mirror. My understanding is that most of the risk faced by life insurers is investment risk, not mortality risk or longevity risk.

    ReplyDelete
  6. Hi guys,

    Thank you for the discussion you are having.

    I'm not an actuary either.

    But I do know that the Social Security Administration in the US does keep track of what is known as period life tables and cohort life tables. What Larry describes sounds like the period life tables. For instance, take the mortality rates at each age in 2007 and use those. But they realize that the mortality rates for 70 year olds in 2007, for instance, do not represent the mortality rates for 40 years in old 2007 for when they become 70 in 30 years. For that, they have the cohort life tables which incorporate expected mortality improvements.

    I would have to think that the insurance companies are making lots of efforts to develop cohort life tables.

    That being said, the actual mortality risk will be different from the expected values. This could introduce new risks, especially if people live longer than expected.

    I was talking to Michael Kitces about this today too, and he made a good point. Investment risk will also be introduced if people live longer than expected, because then the companies will have to rollover investments and the new unknown future interest rates.

    ReplyDelete
  7. Yes, our research actually uses two period life tables, 2007 Social Security and 2000 Annuity. But the dynamics of using the tables is much more than currently perceived; result, static thought leads to static results. My physics background brings in a dynamic perspective to the topic.
    Aaron’s example is JUST the beginning. US Social Security 60 year old male expected longevity is about age 82 under Social Security Tables (age 86 from 2000 Annuity Table – tends to be a healthier population in this table). I’ll use Social Security tables here for the example going forward so there is only a single set of numbers to absorb, but the concept I explain here is the same, the difference simply is what distribution period length results. Okay, so that is a 22 year distribution period in the simulations to determine a prudent withdrawal rate (you need to read our paper series (links to them all accessed by clicking on my name above) to see how withdrawal rates are very sensitive to the length of the distribution periods).
    However, nobody EVER reaches their CURRENT AGE expected longevity age! Expected longevity is ALWAYS older than you currently are … survivors NEVER reach their current expected longevity. So, continuing our male retiree above, when he reaches age 65 his expected longevity is age 83; age 70 it is 84, age 75 it is 86, at 80 it is 88, at 85 it is 90. So the distribution period dynamically adjusts as they age. But, that is using the 2007 table! 20 years from now the table will NOT say an 80 year olds’ expected longevity is 88. It most likely will be older than that under current extending longevity trends, but we don’t know what it will be today.
    So two things are happening simultaneously: First, the distribution period slowly and dynamically changes (lengthens). After 20 years the initial distribution period for our 60 year old male says only 2 more years of income WAS needed. Yet at 80, they may need 8 more years (so it slowly morphed into a 30 year distribution plan, not 22 (and extends into an even longer plan the longer they continue to live). Second, over that same 20 year period the table has extended to some, as of today, unknown value longer than that.
    The effect of a retiree who continues to live (and I submit it is the alive retiree that is important to the adviser as a client than a deceased client) is that what started out as a 22 year distribution period may turn into a 35+ year distribution period with good odds to continue to extend both by his continuing to live, and by changing period life tables. The effect for female and joint tables is the same, just longer respective distribution periods (and lower respective withdrawal rates).
    So the period life tables are a more rational method (rather than ask the retiree how old they think they’ll be some day) to establish distribution periods. Prudent withdrawal rates are very sensitive to the length of the distribution period, so a prudent use of the tables is suggested over guessing how long a period should be. And, both the age from the table, and the table itself change over time! I declare that static distribution periods are out!
    Portfolio values change much more often than distribution periods, so recalculating sustainable and prudent withdrawal rates each year, and recalculating the distribution period by using current tables and current ages, dynamically brings in both sets of probabilities and keeps their plan current. I submit that this is not a set-and-forget exercise when they initially retire. Reviews are required. Our paper series explains the above along with a 3D model that it all fits into. The papers show a glide path for retirees who continue to live into centenarian or super centenarian ages. Withdrawal decisions that are made early on do have a dramatic effect on portfolio balances and sustainable rates at older ages. Taking too much too soon comes at a cost.

    ReplyDelete
  8. To Wade's final point above about his discussion with Kitces ... investment risk. This is precisely the point I am making in the original post Wade made. The growing effort to provide guaranteed income through annuities in 401ks may be the first rung in a set up for too big to fail insurance companies years from now, guaranteeding income to retirees. The solution proposed doesn't address the real problem 401ks have ... not saving enough in the first place.

    The retirement income must remain connected to the asset base generating the retirement income. Otherwise, promised benefits may slowly outpace sustainable income and the pension problems today result. I have no proof, but part of the pension problems may be a static application of expected longevity. Certainly investment results contributed as well. Our paper series suggest how to keep the two firmly in a prudent process.

    ReplyDelete
  9. Re: the concern that Michael mentioned to Wade, unless I'm missing something, the biggest financial hit would be from paying out for a longer period than expected, not from possible interest rate changes.

    I'm somewhat familiar with how actuaries use mortality tables and I am comfortable that they take potential future mortality improvement into account and also put in extra margins. So I'm not worried about insurers' taking undue longevity risk.

    I'm personally more concerned if the current low interest rate environment (including low credit spreads)continues for a long, long time--and I fear that is enough of a possibility to be worth worrying about.

    ReplyDelete
  10. Yes Joe, that is a concern. A withdrawal approach should include a method (and upfront discussion with the retiree beforehand) to retrench spending. A bad market sequence is the first threat that may require a spending adjustment. Continuous poor returns would be another. Both have the same effect on portfolio values ... values go down. The difference is that bad sequence is faster, while extended poor returns is slower (portfolio value goes down from spending effect). Our JFP Nov 2011 paper discusses a method that is forward looking, as opposed to Guyton's backward looking, method to determine when retrenchment should be taken as a serious retiree option.

    Michael Kitces and others have been talking about low interest rate effects. I think we advisers should put an arrow in our arsonals and include it as an upfront discussion with retirees.

    Annuitizing simply shifts the risk from the retiree to the insurance company. Everyone invests in the same markets so market returns are not going to bail out one group (insurers) when they can't bail out the first group (retirees).

    What is a universal truth, success towards either a savings goal or spend down goal is mostly affected by spending and saving ... the main lever that makes a difference. Everything else has either a smaller effect or is out of ability to be controlled, especially over a lifetime.

    In summary, this part of the conversation deals primarily with the probability of the portfolio (sequence risk). The majority of comments up to this point deal with the probability of the person (longevity risk). Both risks need to be considered simultaneously since they both impinge on "prudent" or "sustainable" withdrawal rates (I don't use the term "safe" anymore).

    Great discussion!

    ReplyDelete
  11. This is a good discussion. Re: Larry's comment--"Annuitizing simply shifts the risk from the retiree to the insurance company."--I think a key difference is that the insurance company can pool longevity risks and an individual cannot, so there is quite a difference there. I think of it this way--in an oversimplified no load world, if I'm 65 and want to use a bond ladder to make my retirement work, I need to plan for the bond ladder to last 30 years or so to be safe. With the annuity, I buy into a pool based on a 22 year average lifetime so I can get a higher withdrawal rate that way.

    ReplyDelete
  12. A back of the envelope comparison I've made comparing the two is at this blog http://blog.betterfinancialeducation.com/sustainable-retirement/how-does-an-annuity-compare-to-total-return-retirement-income/ . Bascially, at any given age initially, annuities do provide a slightly higher monthly benefit. But when that is compared over the remaining expected longevity period, the total paid out compares with a less than median market return sequence across the board. However, at later ages, the monthly payment from an annuity, at that age, is lower than what is possible from a total return managed portfolio. Thus, it is not the mmonthly income you can get at the moment (at any specific age), it is what is the range of possible overall potential income you can get over the rest of your lifetime? And, on top of that, what if markets do better, or worse? The lifetime income streams line up with poor markets percentiles in my graph in blog link (copy and paste if necessary) above. With annuities, any market improvements go to the insurer, not the annuitant.

    ReplyDelete

  13. my last comment seemed to not have made it:

    Thank you to all the participants for the great discussion.

    Larry, I really like your term "prudent" withdrawal rate. I recently starting using the term "optimal" withdrawal rate to consider that people might be willing to use something more aggressive than what is safe if they are able to make later cutbacks. Prudent gets at that as well. Safe is not a good term, as nothing is safe about using a volatile portfolio.

    I'm on board with the "mortality-adjusted constant probability of failure" approach you are advocating. That is quite a mouth full. I need to read your articles, but I at least read the Blanchett and Morningstar article on the subject and I think it's gonna be the best way for thinking about a systematic withdrawal rate strategy with some more realistic dynamics to it.

    Thank you, Wade

    ReplyDelete
  14. Larry, if I read your blog post correctly, it looks like you're using historical returns for stocks and bonds in your comparison to immediate annuities? I realize that's a personal choice, but I think it's inconsistent if you compare current immediate annuity payout rates and historical bond returns. (Perhaps I don't correctly understand what you're doing, but that's the impression I got.)

    This describes the assumptions I use: http://advisorperspectives.com/newsletters13/pdfs/Predicting_Asset_Class_Returns-Recommendations_for_Financial_Planners.pdf

    ReplyDelete
  15. Joe, I used our research data generator for the comparisons. That is based on Monte Carlo simulations that generates an Annually Recalculated, Serially Connected, withdrawal amount and portfolio balance each year for the 5th, 25th, 50th, 75th and 95th percentiles of those simulated values (you will need to read the papers at JFP and working papers on SSRN to absorb this … way too much for a blog post here … link to everything by clicking my name).

    I agree with your inconsistency comments, which I interpret as the up-to-now approach to comparison, since traditionally it has been hard to compare apples and oranges. But, with our dynamic process that we have developed and written about, the comparison can be simplified by comparing monthly payouts based on age (compared at the same age) as well as total sum of money received over the remaining expected lifetime (sum of total annuity payments received compared to sum of total Monte Carlo payments received … in this case the 25th, 50th and 75th percentiles were calculated and illustrated on the graph in that blog link; both sums over the same expected longevity time frame from Social Security period life table). This begins to turn the genetics of oranges into apples so an apple to apple comparison is easier.

    I’m toying with the idea of turning this back of the envelope work into a research project so it is more academic … but that’s a matter of having time, getting historical data for immediate annuity payouts, and looking at it from many different historical time frames and angles … not a simple project. However, my initial cut at it, which is in that linked blog posting above (15th @ 7:33AM), appeared sound to my collaborator Dr John Mitchell.

    I was surprised that the annuity payout pretty much consistently aligned with “poor markets” over the remainder of the retiree’s lifetime. However, cracking that tendency open more is what the project I mentioned above needs to do. I think the benefit of the results so far is to bring more light on just what a SPIA actually does for a retiree when you peel everything else away.

    ReplyDelete
  16. Sorry for the typo in the post above ... instead of 15th @ 7:33AM is should read 16th @ 7:33AM.

    ReplyDelete

  17. In my last few articles, including the efficient frontier article in this month's JFP, I've been pushing along with Joe that you can't use historical averages to guide systematic withdrawals while using current annuity prices. Even if looking at a distribution of outcomes, there is no meaning because the entire distribution for systematic withdrawals is way too optimistic. The fact that annuity payouts are low should provide a good clue that sustainable withdrawal rates are also low.

    ReplyDelete
  18. I think one way to look at this is to look at the TIAA-CREF experience because their annuities do not incorporate guarantees. Nevertheless, at least in the past, their mortality assumptions seem to have been a bit conservative. With respect to CREF (mostly stock-indexed annuities), they say "In addition to investment returns, variable payments from CREF vary with mortality and expense performance, although these have had relatively insignificant effects in the past." The payouts from the "fixed" annuities of TIAA (note that these are not guaranteed; they can be adjusted but aim at stability) have actually slightly increased from year to year as reserves are released; they are even increasing a bit next year in the current interest rate environment. So at least in the past, their actuaries have conservatively predicted future mortality (and they have high payouts with respect to the rest of the industry). None of this is a guarantee for the future, of course. But it does suggest that if there is a problem with mortality risk in the future, it is more likely to come from a sudden shock (e.g., a drug that greatly extends the lifespan of those with certain kinds of cancer) than as a result of the usual steady improvement.

    ReplyDelete
  19. Wade, I agree with the distribution of outcomes comment to some extent in that the historical returns used to generate simulations is biased with a tendency for positive returns (the distributions of annual returns graph is skewed a bit towards the positive side).

    However, the current low annuity payouts that I graphed in link above are low based on low interest rates today. I don't think the withdrawal rates are low though as you say; if returns are skewed towards the high end, then the withdrawal rates would also tend to skew high. If this skewed tendency were corrected somehow (lower the returns distributions, this would make the annuity comparison a bit more in favor of the annuity payouts. But keep in mind, currently annuity payouts appear to compare with below median returns by quite a bit.

    This is one of the components that should be academically explored as I mentioned to you. Using a different historical time and data (late 70's or early 80's as data cut off?) to see how such a comparison looks. Then a good historical cross section may emerge. The beauty of research is that it seems to generate more questions and more research to pull out good answers.

    As to anonymous’ point; this appears to be a possible reason annuity payouts may be low too; reserving assets to cover possible trends. But, this leads to lower monthly payouts early on, perhaps "cheating" (not the best term, but a better one escapes me at the moment) the early-to-die in order to provide more income later. Although this might be more honest with a retiree than asking them to retrench spending, I think the choice should be the retiree’s rather than an entity with different motives.

    Finally, I'm in the school of thought that prediction is impossible. One needs to make the best educated estimate possible, and change the estimate as the facts change. In truth, when you look closely at "predictions," this is really what they are doing. I think using the term estimates under the current circumstances is a bit more truthful and believable than calling them predictions.

    I have the deepest respect for actuarial science ... data gathering and statistical analysis is important. The initial point of my comments to Wade though is to beware of the "silver bullet" mentality that a guaranteed income in any form is going to solve the retirement income problem when the asset base gets separated from the promised benefits to be paid (what happened to pensions). Thus our research to get a better handle on how to keep the two inter-connected is both good and bad markets.

    Really great discussion going. Thanks Wade.

    ReplyDelete
  20. Wade says it very well in terms of the relationship between current SPIA payout rates and sustainable withdrawal rates. His efficient income frontier in the Feb. JFP (page 50) shows how stock/SPIA mixes provide a higher level of sustainable withdrawals than the more traditional stock/bond mixes when current SPIA rates and current interest rates are used to provide an apples-to-apples comparison.

    The Anonymous discussion about TIAA/CREF is interesting. I believe the CREF annuities are immediate variable annuities so most of the yearly income changes emanate from investment performance. I think the TIAA "fixed" annuities have both a guaranteed rate and an amount they annually declare above the guarantee. To make things more complicated, the rates differ depending on when the client's money was invested. It's a bit surprising if the annual "bonuses" are increasing in this interest rate environment, so perhaps they have been conservative on longevity estimates. There are so many moving parts that it's hard to tell.

    Re: financial problems with pensions foretelling possible problems for SPIAs, the cases are quite different SPIAs are backed with bonds and companies do duration matching. Pensions have been backed with stock/bond mixes and many plan sponsors ceased making contributions in the heady 1990s and weren't prepared to withstand the first decade of this century.

    ReplyDelete
  21. Good comments Joe. Which is why I'm in the initial developing stages for research comparing what a retiree MAY get under SPIAs to what they MAY get under prudent dynamic withdrawals. At the end of the day, the question is "What is the TOTAL SUM of money a retiree may receive during the remainder of their expected lifetime?" How that money is generated is really immaterial when you get down to brass tacks. For any sustainable income method to be successful there needs to by many broad based markets and many different income systems. If they all converge into one … then it becomes too big to fail.

    PS. my Feb 17th 1:15PM post, last sentence, should read "...in both good and bad markets." My organic spell checker (brain) often fails whith is, it, if and in!

    ReplyDelete
  22. As long as Larry Frank takes the position that prediction is impossible, there seems little point in discussing many of these issues with him. Given that point of view, any success actuaries have had in the past is irrelevant, presumably just due to chance. For example, he now criticizes TIAA ("cheat") because its "fixed" income stream has increased a bit. Think about that for a minute. I'm sure TIAA reserves against unfavorable mortality experience, and I believe that their regulators require that they do so. Suppose for a moment that their actuaries had predicted mortality exactly. Then, as the company observes that mortality is working out as predicted and the number of annuitants decreases, it can release some of those reserves. A for-profit company would presumably take them into profit. TIAA is a non-profit, so it uses them to increase its payout -- and, as thanks, is accused of cheating. They can't win. I did point out in my orginial post that their annuity rates tended to be on the high side in the past, thinking to defuse this criticism. No such luck!

    ReplyDelete
  23. Thanks for the informative discussion. The methodologies of creating the actuarial tables lead me to one more question (which i think I know the answer to). It's one thing to rely on the actuarial tables, but it's another to know what sort of population profile to actually use. I can assume that the age bias of annuity buyers skews towards longevity. There's not a causation, e.g., buying an annuity doesn't mean you'll live longer, but rather, a correlation - people who are terminally ill or not healthy probably don't buy annuities. Do the actuaries account for this selection bias?

    ReplyDelete
  24. Jason, actuaries do take selection bias into account by gathering inter-company mortality experience on specific products and using that as an input to pricing that also requires estimating future mortality trends. So the annuity mortality tables that actuaries use will have longer life expectancies than the life insurance tables. Also actuaries use what is known as select and ultimate mortality, where the select mortality take account of the number of years after purchase, because the effect of selection bias decreases over time.

    That being said, there's lot of judgement that goes into actuarial work. You could get a flavor for that by looking up "Living to 100" which is an ongoing research project involving actuaries and others involved in longevity research. It's not just a "green eyeshade" occupation.

    ReplyDelete
  25. Just adding one thing--the site for "Living to 100" that I was referring to is livingto100.soa.org, there's another livingto100.com, but that's not it.

    ReplyDelete
  26. Let’s separate two topics here. Different perspectives are coloring the conversation. Probability of the Person (longevity and actuarial tables) and Probability of the Portfolio (sequence risk). Anonymous’ comment refers to actuaries which is not what my prediction statement refers to. Market return sequences cannot be predicted, thus future portfolio values are unknown. That is why Monte Carlo tools have been used. However, probability distributions and use of percentiles are not the same as predictions; they provide insight and nothing more. The period life tables used today are based on data gathered from the population that lives versus the population (cohorts) that died. There are no cohort tables developed yet for people born in 2014 or later. The period life tables for cohorts don’t include any predictions. Our research looked at both the 2000 Annuity table and the Social Security table. Yes, the annuity table did have a slightly older expected longevity (4 year difference at age 60) than the Social Security general population table. However, by age 100 the differences between tables disappear. Those differences do have an effect on prudent withdrawal rates based on age profiles. Joe’s website reference is a great resource … and our JFP Dec 2012 paper explains what happens to retirement resources and the withdrawal rate glide path when a retiree lives to 100 and beyond.
    We are now getting into details that go off track from the theme of my original email to Wade … and missing the larger picture. Defending the status quo is how the mortgage business and subsequent investment banking got into trouble. My comments to Wade were merely to bring into point of discussion a tendency to promise more than the system can bear. There is only one global market that everyone invests in … pensions, insurance companies, endowments, foundations, and everyone else. Good times, normal times, and bad times affect all participants the same in the end. Benefits should be kept in line with what the markets provide, otherwise more may be promised than the system can provide. I merely suggest that is what should be considered.

    ReplyDelete
  27. I tried a post, but I hit the wrong key and it went into never-never land, so I'll try a shorter version. Just three comments--insurance companies often customize mortality tables based on their own experience and include mortality improvement projections even if they are not included in the base table like the Annuity 2000 table. I think actuaries have a pretty good handle on longevity assumptions. A second comment is that, when raising concerns about risks, one has to also consider the amount of capital being set aside to cushion the risks. Talking about risks without also discussing capital adequacy is only looking at half the picture.

    ReplyDelete
  28. Thanks everyone for the continued discussion.

    Jason, you do bring up some good questions. Annuity purchasers do tend to live longer than the average population. This could be because both adverse selection (those with some knowledge that they may live longer than average are more likely to buy the annuity) and because qualities important to long-term financial planning such as education and wealth tend to also be correlated with longer life spans. This is all correlation. And surely, as Joe says, actuaries make a lot of effort to stay on top of this so that their employers do not go out of business.

    I have seen some claims about causation as well, though I'm not sure whether they are the product of the overactive imaginations of annuity salesmen, or because they're based on science. To test this would require some sort of experiment where people are randomly assigned whether or not they receive an annuity. But the claim is that annuitization causes peace of mind and reduced stress about one's finances, and this in turn leads to a longer life. Again I'm not sure about the science behind this.

    Overall, I think Larry has brought up an important point that we need to keep in mind. At the societal level, there are potential unintended consequences that could result from getting more and more people to annuitize their assets. This is something to be mindful about.

    ReplyDelete
  29. quick two cents. It's a bit unfair to compare SPIA payout rates today today with long-term return averages. A better approach is to incorporate a return model that more accurately reflects today's available yields for investors (especially bond investors). Here are two papers I wrote with Wade and Michael Finke that move in this direction.

    http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2201323

    http://corporate.morningstar.com/ib/documents/MethodologyDocuments/ResearchPapers/LowBondYieldsWithdrawalRates.pdf

    The next step is obviously to roll this analysis into the optimal annuitization decisions, where both the withdrawal amount and annuity decision are dynamic. These are works in progress!

    ReplyDelete
    Replies
    1. Yes, this sort of dynamic returns model is the key to being able to properly analyze both the decisions for when to annuitization and whether to use a buckets approach for retirement.

      Delete
  30. Agreed David on the surface. Another approach is to look at data from various historical periods when comparing the two. Ultimately, comparing the total paid to a retiree over any period of time from both income methodologies equates the comparisons too. Total dollars received is the common denominator in this concept. Granted, that is a historical look, however it would be a start for comparison and insight. It's not going to be fleshed out here in an exchange of comments though ... this needs research work. As you say, a work in progress...the beauty of research, there's always a question to answer and more to understand!

    ReplyDelete
  31. Investment leads provided by a trusted and reliable supplier. We currently supply a number of investment brokerages around the world with their investor leads. Take a look around and see how we generate the investor leads.
    investor leads

    ReplyDelete
  32. Thanks for adding up my knowledge and approach towards annuity decision making. There are million factors that affect it but one simply can't ignore 'viability'.

    ReplyDelete
  33. Tax avoidance is the legitimate reduction of tax through tax planning or usage of legal provisions. Unlike most other countries, most UK tax professionals are accountants rather than lawyers by training.
    company tax avoidance

    ReplyDelete
  34. Employee benefit trusts allow you to withdraw profits from a business and pay less tax in comparison to traditional salary or dividend payouts.
    employee benefit trusts

    ReplyDelete
  35. It is important for small business owners to maintain a personal awareness of tax planning issues in order to save money.
    Visit here

    ReplyDelete
  36. You want to Get Pension Release. Looking for the Pension Release Services can help with general guidance about pensions release advice in UK, get more services like, Get Pension Release, Pension Release, Pension Release Services.

    ReplyDelete