Saturday, December 29, 2018

Lessons from 2018

Retirement finance is my hobby. I find it has two great rewards: helping people who can’t find affordable retirement advice and learning new stuff.

OK, sometimes there are other rewards. A retiring college professor insisted on paying me for her retirement plan so I negotiated dinner with her at a new pizza place I had wanted to try. (Yes, I work for pizza.) A podcast producer interviewed me and afterward sent a really nice set of engraved kitchen knives. My wife is an amazing cook. She is also an MBA but opening those knives was the first time in two decades that she has shown any interest whatsoever in my retirement planning hobby.

Back to learning stuff, I’ll wrap up the year by sharing a few of the things I learned in 2018 that you might find useful.

I spent most of this year co-authoring research with Neville Frances, a UNC econometrician. The first thing I learned was that a research project can take four times as long as you expect and dramatically impact the amount of time you have to write blog posts. I hope to do much better next year.

I learned that a lot of poor research is published. I was already aware that a lot of excellent research is available in the retirement finance field but also a lot that is questionable. When I discussed this with my co-author, he shrugged and told me that’s true of all of economics. So, I checked with my son, a medical researcher. He confirmed that he runs across a lot of junk in that field, too.

If you read something in a peer-reviewed journal, be skeptical. If you read it elsewhere, be very skeptical. I’m planning a column on the topic for early next year but Francois Gadenne has already published his thoughts.[1]

Most papers are essentially arguments. I find one of the problems with reading papers is that many people can’t logically deconstruct an argument. I read a paper this year, for example, that made several claims but provided no evidence to support any of them. If the basic argument is flawed then discount the research.

If you’re interested in analyzing arguments, my friend, Dr. Walter Sinnott-Armstrong, created an excellent, free video class at Duke University[2] and he has written a couple of books on the topic with Think Again perhaps the more readable.[3,4]

I learned the extent to which sustainable withdrawal rate (SWR) is explained by sequence of returns as opposed to the portfolio returns, themselves. "Big Ern" at found a convincing way to explain it.[5] He found that "knowing only the average returns over the next 30 years is not very informative."

But he also found that for a 30-year retirement, nearly all (close to 96%) of the variation in the sustainable withdrawal rate is explained by the average returns of six five-year windows. The average return for years 0 to 5 explains about 29% of SWR variance and the average return for years 5 to 10 explains another 19%. Explanatory power declines further in subsequent windows and totals about 0.96.

(If terms like "regression testing" and "R-squared" don't frighten you away, I highly recommend the post or any other on his blog.)

The simple takeaway here is that when you spend from a volatile portfolio, the long-term returns matter very little compared to the sequence of those returns. Don't worry about whether your portfolio will earn an average 8% a year but about when the bad years will occur. (Later is better.)

This doesn't mean, however, that once we survive the first five years of retirement sequence risk goes away.

I used Ern’s spreadsheet to estimate that returns for the first four years of a retirement with 20 years remaining have about the same explanatory power as the returns for the first five years when 30 years remain. The explanatory power of the returns for the first two years of a retirement with 10 years remaining is about the same as these. Sequence risk becomes “compressed” but it never goes away entirely.

If we successfully navigate the first five years of portfolio returns then we still have to negotiate the next five and eventually the next two. There's no reason to expect that if you make it through the first 5 years of retirement that your risk will simply disappear.

I learned that many Monte Carlo models are poorly designed. I also learned that many advisors who use them don’t really understand the technique and by the time they explain the results to the typical client, most of its value is lost. Although I still consider MC an extremely valuable tool, I'm now cautious about recommending it because I'm not confident that it will be used and interpreted correctly.

I learned from Zvi Bodie that probability of ruin (or probability of shortfall) is a problematic metric not only because it measures the probability of a shortfall while ignoring the magnitude of the loss but also because it ignores utility. He explained this in terms of Arrow-Debreu contingent claim state prices, which probably makes as much sense to my readers as it did to me when I first read his explanation.[6] I had to learn about Arrow-Debreu before I could even have a discussion. A-D won't interest normal people but I found it pretty exciting. It even gave me some insight into the Black-Scholes model.

Having less confidence when I retired would have served me well.
[Tweet this]

Simultaneously, I was learning a great deal about MC from my econometrician co-author and it became clear from our work that using MC to measure probability of ruin has even more problems than using the historical probability of ruin. (Probability of ruin isn't a "robust" model metric.)  We don't understand the process that determines the sequences and we don't have enough independent 30-year historical sequences to provide insight, so we don't really know how to model sequence of returns. In turn, I no longer have confidence in analyses that measure Monte Carlo model results using probability of ruin as the metric.

I learned that there is no consensus among economists as to whether stock returns mean-revert or, even if they do, if that would imply that stocks become safer the longer we hold them. (See Mean Reversion of Equity Returns and Retirement Planning.)

Lastly, I learned that I'm not the only one concerned about retirement planning for unwealthy households. Wade Pfau, Steve Vernon and Joe Tomlinson addressed this problem in a hefty tome entitled, "Optimizing Retirement Income by Integrating Retirement Plans, IRAs and Home Equity." You might want to read Vernon's shorter discussion of part of this work referred to as the "Spend Safely in Retirement Strategy" here[7] in which he concludes:
The "Spend Safely in Retirement" strategy represents a straightforward way for middle-income workers with between $100,000 and $1 million in savings to generate a stream of lifetime retirement income without purchasing an annuity and without significant involvement from financial advisers. This group might represent as many as half of all workers age 55 and older.  
This is hardly an ideal retirement finance strategy but retirees with limited resources and no access to a good financial adviser might have difficulty finding a better one and I think that was the goal.

So, that's some of the important stuff I learned in 2018. Ironically, the more I learn about retirement finance the less certain I become about what I thought I knew. Being less confident back when I retired would have served me well.

Wishing you a happy and enlightening 2019!


[1] How to Read Research Papers With a Discerning Eye: Take the Best and Leave the Rest by Francois Gadenne.

[2] Think Again I: How to Understand Arguments, Coursera.

[3] Understanding Arguments by Walter Sinnott-Armstrong.

[4] Think Again by Walter Sinnott-Armstrong.

[5] The Ultimate Guide to Safe Withdrawal Rates – Part 15: More Thoughts on Sequence of Return Risk,

[6] An Analysis of Investment Advice to Retirement Plan Participants by Zvi Bodie. (see footnote 10)

[7] Meet the "Safe Spending in Retirement Strategy" by Steve Vernon.


  1. Dirk, thanks for sharing your hobby and learningswith us!

    It's interesting that Vernon's recommended "Spend Safely" strategy is "without purchasing an annuity". I'm thinking more retirement income can created from a given portion of savings by purchasing an annuity (SPIA) than a safe spending rate from that same amount of savings in a portfolio. This is because mortality credits are taken advantage of. (correct me if I'm wrong here). However It could be Vernon is giving heavy weight to the annuity disadvantages of you can't easily change your mind, and complexity of some annuities.

  2. A fixed annuity will probably provide more income than say, the 4% Rule, initially but portfolio spending will provide more income if the portfolio grows and less if it shrinks. We can't know for sure which will provide more lifetime income because we can't predict portfolio returns. The annuity will provide more lifetime income if the portfolio shrinks, so it provides downside protection (it's purpose), but will provide less lifetime income if the portfolio grows (its downside).

    To your point, a fixed annuity for a 65-year old couple right now pays out about 5.5%, significantly more than the 4% Rule, for example. Whether it continues to pay more than the 4% Rule depends on the market. Also keep in mind that the "sustainable withdrawal rate" increases as you age.

    Bottom line, you have to ask yourself whether your willing to risk some of your standard of living in hopes of improving it with SWR, or guarantee a floor while giving up the hope of spending more. It depends on your risk tolerance to a large extent.

    Perhaps Steve will jump in here but I don't thinking he is making a judgment regarding annuities. I think he is acknowledging the fact that most retirees, rightly or wrongly, simply won't buy an annuity. He's saying, "This is a pretty good strategy if you a) have saved between $100K and $1M, b) don't want to work with an adviser or can't afford one, and c) don't want to buy an annuity."

    You sound like you're not in the target market for the strategy because you are willing to consider an annuity.

    Great question. Thanks for writing!

  3. This was an interesting column as usual.

    Per the quality of technical papers, here is how I classify technical papers as somebody who has written some and used a lot of the years:

    1. State-of-the-art or state-of-the-practice summary that is very useful in understanding the theory and defining good practices. Usually written by a major player in the field.
    2. Re-defining paper - a new or significantly revised body of knowledge paper that is a significant addition to the field. Usually written by a major player, but not always.
    3. Data point paper - good experiment or case history, well-thought out, with good documentation, usually very useful when people do state-of-the-art summaries or redefining papers. Can be written by anybody with interesting work who did well in high school science class.
    4. Repeat paper - there is no point in doing a research project without getting at least 5 papers on the subject, preferably with funding to attend conferences. Find the latest one -ignore the rest.
    5. WTF? - no further discussion required.

    I pay a lot of attention to the reference lists in the papers. They are a window into the soul of the writer. A good reference list shows that the writer has a firm understanding of the fundamentals of the field and understands the various schools of thought. A bad reference list generally means that the point-of-view of the writer was pre-determined before he even laid out the goals and objectives of the paper.

    On reversion to mean of stocks: I think this a generally valid concept on the basic understanding that the general trend of the country and economy is up. This has been true for the US and Canada over the past couple of centuries, but not so certain for many other countries. So I view it as an assumption where it is unclear what the alternative would be, other than owning guns and farmland.

    If you look at 30 year swaths, we have 1990-2020 (almost), 1960-1990, 1930-1960, 1900-1930. Each of these periods is largely entirely unlike the other in many ways, yet historically things have typically gotten better over those periods.

    There is a pretty good data set from about 1970 to now for a wide array of assets, including international. This period includes the 1970s inflation and crushing interest rates, the huge bond bull market with declining inflation, a major financial crisis, the fall of Communism, etc. I go under the working assumption that a portfolio that does acceptably over 10, 20, and 30 year periods since 1970, that it will probably do fairly well in future market conditions. I haven't figured out why Monte Carlo would give me a more reliable analysis than this although it looks fancier.

  4. I read your recent article for Forbes and wanted to tell you that I appreciated your honesty with how you prepared and started your journey into retirement.

    It struck me because of how similar your situation is to so many individuals and couples I see preparing for retirement in the shifting sands of the market today. What would you say was the reason you didn’t look into any of the things you described in your article? Was it lack of education and exposure to that information or was it because you believed that higher investment return would weather any storm?

    I deal with this problem everyday as I attempt to explain to my clients the importance of LTC planning, investing prudently with a portfolio objective based on their risk tolerance and time horizon, as well as having 2 years in cash or cash equivalents to navigate a bear market.

    I’d love to hear your thoughts and see what you would recommend I do to convey the importance of these planning details.

    1. If I assume that "lack of education and exposure to that information" are essentially the same thing, and I limit "lack of education" to lack of education specific to retirement finance then, yes, it was primarily due to lack of education. I would quickly add that I started with a degree in computer science and graduate level training in statistics and economics.

      It didn't help that people who sell stocks advertise and write blog posts with simple answers while economists who study risk publish dense articles in inaccessible journals.

      The question you ask — a good one — is how your clients can avoid my mistake. Education is an obvious candidate but retirement finance is complex and few of your clients are likely to start with the foundation I had. It has been challenging for me even with that foundation.

      Some clients will readily admit that they have a limited understanding of these complexities. Others convince themselves that they understand when they clearly do not. In either case, to make decisions in a highly complex field we have to become educated (a long, tedious process for retirement finance) or decide which expert to trust. There are many such "experts" to choose from, many with broadly differing opinions.

      You are a retirement planner so I suspect that the best answer is to make them trust you as that expert. That will probably mean developing confident, credible, understandable explanations for why risk tolerance, for example, is important. But educating them will have diminishing returns as most don't want to become experts, themselves.

      Otherwise, they will seek a quick, easy solution like "buy stocks and spend 4%", whether or not that is a good strategy but because it is a simple strategy.

      From the approach to planning you describe, I see no reason for them not to trust you.