Thursday, August 16, 2012


I'm back from a brief trip to the Boston area, where I spent a couple of days at the Retirement Management Analyst (RMA) intensive seminar at Salem State University. The Salem/Marblehead area is a beautiful part of the country and full of colonial history. And the seminar participants were all very interesting and came from a diverse set of backgrounds in the financial services world.

On Tuesday, I was presenting at the seminar for most of the day (I think I was talking for about 5 hours in total, using a presentation that I can get through in one hour when pressed, as there was lots of good accompanying questions and discussions), but I also enjoyed the explanation by one of the participants about software used in practice which follows Moshe Milevsky's product allocation approach.

On Wednesday, the special event was the visit by Laurence Kotlikoff, who presented about his ESPlanner software for lifetime financial planning. I actually became a paying customer for that software back in 2007, but then had gradually gravitated toward a spreadsheet I had made for my own household's lifetime financial planning. Nonetheless, it was interesting to hear Prof. Kotlikoff's explanations about how the software was constructed, which gave me a better understanding and made me interested to have another look at the software, if only to confirm that I get the same general results as with my own spreadsheet.

ESPlanner is based on underlying principles of academic economics. It is based on lifecycle finance, which involves maximization of lifetime expected utility. However, very cleverly, Kotlikoff built the software to simplify a few steps away from formal utility maximization. That is because using a formal utility function involves the unrealistic expectation that we can define such variables as a person's risk aversion coefficient, minimal consumption floor, a factor for how they view the tradeoff between spending today and spending tomorrow, and so on. What Kotlikoff does instead is allow users to specify whether they wish to make spending plans based on the assumption that they will earn the expected return from their investment portfolio, half of the expected return, a real return of zero, or an upside approach which assumes that the value of all stock holdings will go to zero, but will then add upside spending later as stocks are converted to TIPS. That last one was inspired from talking with Zvi Bodie. This is an alternative way of allowing users to estimate their own attitudes toward risk.

And we must be clear about what risk means in ESPlanner. As ESPlanner is based on the economics approach to lifetime financial planning, what people want to do is to smooth their lifetime consumption (basically, enjoy the same living standard for each member of the household for each year of life through a maximum planning age). People want to smooth their consumption and to get that smoothed consumption level as high of as possible. And so the risk is risk to one's lifestyle. That is the risk that events will take place which force consumption to decline from the currently determined level.

Lifecycle finance manages this risk by shifting consumption from good states of the world to bad states of the world. In technical terms, consumption is shifted from low marginal utility states to high marginal utility states. In other words, when things are working out well you feel more satiated and will be happier on an ex ante basis by shifting some of that consumption to the bad luck cases when consumption is low. That is when a bit more consumption would be gladly welcomed.

Risks to lifestyle are managed through precautionary savings (save more than needed in case you lose your job or experience bad market returns), diversification (by diversifying between stocks and bonds you lose some upside when stocks go up but avoid some downside when stocks go down), insurance (decrease consumption by spending on insurance premiums in good states of the world which pay off with protections in bad states of the world), and hedging (find strategies where bad outcomes are offset with a good outcome from the hedge and vice versa).

Though the point is quite obvious, there is something Prof. Kotlikoff said that resonated with me as being a really important way of stating the matter. That is, risks to lifestyle depend not just on investment decisions, but also on spending decisions. The more aggressively one spends, the greater the risk to future lifestyle. I'm still thinking through whether aggressiveness in spending will also be linked with aggressiveness in investing, or whether a person can rationally be more aggressive in one aspect but less aggressive in the other. Perhaps these are two sides of the same coin. As I think about it, I believe this was one of the results we determined in the "spending flexibility" article with Michael Finke and Duncan Williams. I need to read that again in light of what Larry Kotlikoff said.

Do note as well that an important point about consumption smoothing is that it tends to throw out the window the traditional notion that people should aim to replace 80% of their gross salary in retirement. That is a rule of thumb that will only be optimal (even approximately) by coincidence. Replacement rates should properly be based on what smoothes one's living standard over time, both before and after retirement. This basic rule of thumb could cause some people to over save and defer too much pre-retirement consumption, and it could cause others with a large amount of financial assets relative to income to spend much less than necessary in retirement.

The software is very interesting. When I used it 5 years ago, I guess I gravitated toward my own spreadsheet because ESPlanner only considers one scenario at a time, and to compare different strategies (different asset allocations, investing in different types of tax-sheltered vehicles, etc.), it required re-running scenarios after changing various assumptions. It would also require re-running things to consider changing assumptions such as the ratio of spending needs of children compared to adults (one might play with assumptions such as that children only spend 70% as much as an adults) or how much more spending is needed to support two adults living together, compared to one. That is also true with my spreadsheet, but because I do not need many of the assumption screens built into the software for my more basic personal situation, it was easier to keep track and change assumptions in a more basic spreadsheet in which I smooth consumption manually rather than with dynamic optimization routines which Kotklikoff developed. The basic idea of what I aim to do in my own planning spreadsheet does match the underlying approach I learned from ESPlanner. At any rate, I am eager to have another look at the software. A relatively sophisticated free basic version is available at his website.


  1. I understand that we want to smooth utility over time, which, implicitly, means smoothing income. However, as Chris Christensen so poignantly points out (, there comes a time when we can no longer physically do certain things and slow down our material/experiential consumption, perhaps to be replaced by healthcare consumption. After all, if I have difficulty walking, then a long trip is unappealing, and if I can no longer drive, then the nice car won't do me any good. Did you find the software accounted for this expected shift?

    Additionally, if, as humans, we're used to discounting future utility such as Thaler posits, then are we going to be prone to spend more aggressively now. In general, we are willing to subject the future self to less pleasure in exchange for current pleasure; it's why we join gyms but always say we'll go there tomorrow. Does the software make it difficult for us to fall victim to our own bias towards our present utility?

    I'm a fan of the LCSI approach, but also believe that we can't just blindly rely on modeling lest we let our psychological biases get in the way of achieving life outcomes which reflect the model outcomes.

    1. Jason, good questions!

      About the first one, I was concerned about that as well. The baseline assumption is that you wish to maintain the same spending until your maximum planning age. But there is an option where you can set the relative spending at different ages to be different. So you can manually adjust the software to account for this issue as you wish.

      About discounting future spending, the software implicitly assumes you value consumption the same at all future ages. But with consumption smoothing, you still end up adjusting spending from year to year if your portfolio includes volatile assets. It depends on market performance. The way someone would deal with this is through the choice about how aggressive they wish to be with their spending, with more aggressive spending leading to a greater likelihood of future cuts. The implications of these assumptions can be explored with the Monte Carlo features in the software. The software does make it easy for you to develop a lifetime plan that if followed will make you less likely to fall victim to this behavioral problem.

      Thank you again, Wade

  2. Wade, I was about to take a new look at ESPlanner as you suggested. But I’m so appalled by the aspect of it described in your paragraph beginning “ESPlanner is based on underlying principles of academic economics” that I’d prefer to eat broccoli. I suspect the latter would be better for my wealth as well as my health.

    I continue to scream that the #1 responsibility of advisors and advisory tools is to do their best to INFORM the investor of goal probabilities and uncertainties of alternatives, so the investor can make informed choices. For over two decades, academics have taught and promoted the opposite: first get the investor choice, then apply the academics' math to see the implications of the investor's choice for his future goals.

    It’s the equivalent of changing “Ready, aim, fire” so the “fire” comes first.

    It’s an easy matter to show the investor how alternatives compare in probabilistic assessments for his goals, so he can make informed choices. Instead, for the last two decades devices such as “risk questionnaires” and “utility functions” have been used to extract key decisions from investors based on technical specs for single-year return rates, without informing the investors of implications of those decisions for their goals. It appears the primary purpose is not to inform and best serve the investor, but rather to facilitate application of the academics’ math.

    It sure appears to me that, as described, the ESPlanner step of asking the investor to choose his return rate is just another way to facilitate application of academic math, more than to best inform and serve investors. If I am asked to make that choice, I answer “No! I know future return rates are uncertain. I want you to SHOW ME HOW THE CHOICES COMPARE IN RESULT PROBABILITIES FOR MY GOALS, so I can make INFORMED choices!

    Am I misinterpreting, and misrepresenting ESPlanner??

    Dick Purcell


  3. Hi Dick,

    In this case, I'd say you are misrepresenting ESPlanner. Just click on the Monte Carlo feature and you can see the probability distributions for what is your #1 goal: your sustainable current and future lifestyle as represented by the amount you are able to spend.

    As I mentioned, my _own_ financial planning is based on a spreadsheet inspired by the ESPlanner approach. It's not a case of math for math's sake.

  4. Wade --

    Thanks for your clear and concise reply.

    I guess I have misinterpreted the paragraph I cited.

    But at the ESPlanner site, it appears to me that to get the Monte Carlo I have to pay, and it appears the Monte Carlo may be done at the end of the process, and if the latter is true I do not want to pay.

    My concern can be resolved by answering this question: Is the Monte Carlo done before or after the user is asked the return-rate question?

    Dick Purcell

    1. Dick,

      I won't be able to confirm this until I get home and get ESPlanner up and running again, but I do believe that Monte Carlo is activated by clicking a button so that the first set of results a user sees include the distributions of spending outcomes.

      Without Monte Carlo, the value of ESPlanner is to provide a lifetime set of spending, saving, and insuring decisions that incorporates borrowing constraints, taxes (federal and state), mortality rates, etc. With the spending, how aggressively you set your spending determines how likely you are to have to cut spending later. It is the spending aggressiveness question which provides the only link back to utility maximization, so no need to specify any coefficient values.

    2. Wade –

      It still appears to me that this ESPlanner just isn’t up to the quality and responsible investor’s-best-interest approach of your work.

      Your work features recognition of, and informing investors to cope with, investment-performance uncertainties and informed investment selection, as well as other major considerations such as saving and spending rates. Everything I’m seeing about ESPlanner indicates it’s just another approach that fails to deal responsibly with the big investment questions, instead making assumptions about investment as a basis for lots of academic math on lesser matters.

      It’s certainly true that magnitudes of saving and spending are most important, as your analyses show. But fine-tuning of those factors is dwarfed by questions of investment uncertainty and investment selection.

      The responsible approach is to inform the investor of various portfolios’ prospects and uncertainties BEFORE selection – and then incorporate the selection and its uncertainties in lesser refinement of the rest of the plan. It appears to me ESPlanner does the opposite – shooting first, then aiming.

      If I remember correctly, over at Bogleheads mathematician Rodc did lots of that stuff, with ESPlanner and with his own modeling. And when he considered investment uncertainties, he found that fine-tuning a life plan was like rearranging the deck chairs while ignoring the big questions posed by nearby icebergs.

      Dick Purcell


    3. Hi Dick,

      Of course it is a great complement for you to say that ESPlanner is not up to my standards. But I'm not so sure, as I still think it is pretty good and far beyond what I can do, as it incorporates taxes, life insurance, and many other features that represents hundreds or thousands of labor hours.

      You make some interesting points though. I know how to program Monte Carlo simulations with distributions of outcomes, but I'm still not completely comfortable with them. For myself, I'd still rather look at the outcomes with a set of fairly conservative return assumptions. I've got a full financial plan through 105. I know that there are many unknown factors and guesses that will cause deviations to the plan as the years pass by, but I hope that with relatively conservative assumptions, the odds are that things will end up working out better than planned. They could work out worse. And I could make some sort of probability distribution for those outcomes. But at the end of the day, I don't find that to be so helpful to my personal planning.

      Where ESPlanner still needs work is to better allow a comparison between different strategies. I guess that is the point you are focusing on. Not much attention is paid to the asset allocation decision. And there is no attempt to optimize with that regard.

      But I think they just haven't gotten that far along yet. I think they'll get there one day. The underlying engine behind what they are doing seems pretty sound to me. This just may be a matter of personal preferences for how you like to see results displayed.

      Though I became frustrated enough with the process of inputting data and comparing results that I just gave up and made my own spreadsheet, as I've said, that spreadsheet is based on the same principles.

      One of these days I make a version of the spreadsheet with personal details removed and post it.

  5. > A relatively sophisticated free basic version is available

    Can't find it on the site


    1. Ah, I agree that is was rather well hidden.

      It's at: