Oct 22, 2017

What I use in my own process - draft 1

[Note: I say draft 1 because I know the second I press the publish button, things will change.  I say process because that's what retirement is.]

Back in July (2017) I started to write a post about the story of how I ended up in a semi-voluntary early retirement at age 50 which is part of the backstory on why I do this blog at all. That was fun for a while but after about 7000 words and 14 pages of unstructured nonsense even I started to lose interest.  I couldn't imagine anyone else slogging through it for more than a page.  While the story itself is pretty good if told right, and could maybe fall somewhere between tabloid grocery-store-style gossip on the one hand or maybe Dostoevskian-like pathos (I'm thinkin', though, that maybe bathos is a better word) on the other hand -- and keep in mind, this is a story that probably should be told at some point -- I still had two problems:  one, most of the protagonists of my story are still alive so maybe now is not the right time to open up on that front and two, I lost track of the main reason I was creating the post in the first place.   The main reason at that time was to delineate, since I am trying to write a retirement finance blog, the main tools I personally use to navigate this stage of retirement (before age 65) that I have been in for a while.  I was assuming that that might be useful to someone else, but in the end who knows?  We all have different situations, capabilities, and inclinations.   

But, let's give it a shot.  Here, then, in short form, or at least shorter form than I first tried, are the main things I bring to bear on my own situation, which at age 59 is starting to look a little more like a normal retirement rather than what felt like a too-early retirement at age 50 when I started back in 2008-2009: 

1. The foundation: a deep sense of personal responsibility and accountability. Ok, so this is not software or math or a tool but it's important.  I have tried to keep what I do grounded in the knowledge that I retired for a purpose: to care for school age children in the midst of an interstate move, a divorce, an economic crash (07-08), and a situation where one parent traveled a lot.  Continuity and presence were what mattered  above all else and the "retirement" served the purpose. That purpose drove almost all of my planning up to this point though I can now see it starting to change in ways I have not totally gotten my arms around yet.  To warp Maslow's hierarchy, the playing field of my early retirement was based on a hierarchy that might look like this: presence, continuity, sustainability, prosperity, legacy -- with the left being way more important than the right.   Almost all decisions were based on that premise. This doesn't look like a "tool" but it really is. A touchstone is better than a spreadsheet or an equation in the end. The "why" can drive everything.  But watch out, the "why" can change and mine is doing just that right now.  Call it a "why" or purpose or whatever you want but I still recommend having one whatever you call it.  

2. Blank slate continuous time thinking and a bias for adaptation.  I am permanently attached to nothing when it comes to planning except the goal and even that might change. I use whatever makes sense and whatever makes sense depends on what things look like each day I wake up.  Each morning everything may have changed and I will face a brand new retirement: return expectations mutate, spending expectations shift, inflation comes and goes, asset values?, wars, rumors of wars, the nature of retirement finance itself, etc, etc.  Each morning is a new retirement that requires new thinking and analysis.  This is my problem with things like the 4% rule.  It stays anchored in a "present" that no longer exists.  If the wind changes, using a sailing metaphor, sails and tack will change even though the end goal hasn't moved.  Note this is not even necessarily about things like decision rules, this is about a philosophical posture that assumes the world starts anew in each moment and needs to be thought through again each day.  Of course I might actually stretch out my own do-over thinking to months, quarters or years but that's just because I am lazy. 

3. A strategic plan.  Ok, that's a little bit of a head fake but I at least aspire to having a formal strategic plan.  And here I mean the very simplest of things.  A simple strategic plan plays out the future in a small number of scenarios and anticipates what action will be taken if that scenario-future unfolds and assumes that some rudimentary preparation has been made for taking the action.  A simple example might be hurricane planning. Scenarios might be: hit by none, hit by 1 cat1, hit by 1 cat2, hit by….[you get it]… hit by 2 or more cat1, hit by 2 or more cat2... In each of those scenarios, there are sub scenarios: gas/no gas, social unrest/no social unrest, etc. A strategic plan* would have not only supplies laid away but a course of action depending on the scenario combinations that unfold.  Same for retirement. What scenarios? I don't know, let's say: fast market crash, slow recession, high inflation, low inflation, war/deprivation/refugee, high economic growth, stagflation, whatever…and all of that in combinations.  Me? I know if my net-worth halves for more than 1-2 quarters, I will precisely cut spending exactly where I know I'm going to cut it and I know by how much. Maybe that doesn't look like a strategic plan but it really is in a semi-weak form.   For me I probably need to write it down at some point but at least I have started.

* actually this is more of a contingency plan.  The strategic plan, if I were to have one, would have this kind of thing as well as a game plan for moving from left to right in my retirement hierarchy. 

4. The Income Statement.  I can't back this up with research, which I know exists, but spending, in my experience and reading, is the single biggest lever that non-working retirees (and non-retirees for that matter) have to affect outcomes.  If someone were to be reading a post like this or worried about retirement but not tracking income and spending that is almost completely incomprehensible to me.  This is one of those sine qua non things.  Must do.  Must do first. There is no way to budget or evaluate or plan or improve or adapt if one does not have a handle on this kind of stuff.  And we need both sides of the income statement.  Dirk Cotton: "The expense side will always be uncertain and our goal in retirement isn't really to secure income but to secure non-negative cash flow."  Hard to do that without managing an income statement.  Or here is Ben Carlson: "It’s pointless to try to figure out how much you’ll need in savings or income if you don’t have a good understanding of how much it costs for you to live."   Create an income statement. 

Now, I am not so dumb as to assume that no retiree has an income statement (they do, but not everyone; some think it's too much trouble) or that one can't cut spending without historical data on spending (they can),  it's just that an organized data set can improve the speed and precision of changes to an adaptive spending program if one is embarked on that path.  It also allows proactive planning. For example, if I have a 50/50 allocation and one of those 50s were to go down by 30% and I have a ~3% spend plan with a provisional maximum of 5% in any given environment, then I know not only how much I need to cut spending but I have already planned and identified specifically what goes first.  I'd rather know and act proactively than shoot from the hip a half year after or two after I am already underwater.

5. The Spending Process Control Chart.  Now, one might think that a guy blogging on retirement finance would have some super-secret powerful spending rule kind of thing going on. But one would be wrong. First of all I haven't spent much time figuring out these spend rule approaches though I know they are out there and I know they can dampen risk (I really should look, shouldn't I?). Then, from what little I know about them, even though they are designed to be flexible they, counter-intuitively, strike me as actually a little rigid and they violate in some way I can't yet articulate the principles I mentioned in item #2 above.  Finally, I have to acknowledge that spending is not a fixed thing or even a decision-rule-variable thing it is a random variable.  That means that my own preference -- given that I understand that spend rates: a) change with age, b) are randomly variable and to some extent out of my control, and c) I know that high spending is risky and low is self-depriving -- is that I treat spending more like a statistical process control thing where I plan, execute, measure, and improve (and repeat) continuously over time.  This approach is probably a function of my background in continuous software development processes way back when. 

I start with a spend rate that represents something around a  90-95% success rate level (age specific) that I determine from sustainability analysis (below in section #6). To this I add, of course, some judgement. That becomes my budget for that year and age.  A half percent up becomes my upper control limit (UCL) and a half percent below is my lower limit (LCL).  I then keep a chart with x-axis = month, and y = spend as a percent of "net monitizable assets" (these are assets I can realistically spend in the future net of any liabilities). The line I track in the chart is my monthly spend in % terms. (I also have a 12mo. simple moving average as well as 2 x 2std deviation lines running off of the SMA to either side.)  I overlay the UCL and LCL on the chart, too, but I will update that over time as I get older and my budget and risk changes.  Then I sit back and watch and I adjust things if the process is: a) trending the wrong way and getting too close to the limits, or b) getting too volatile (call this sequence of spending risk similar to sequence of returns risk).  i.e., I keep it under control and then optimize it for whatever I am trying to do.  Even though this kind of thing is a vanilla Demming-style industrial process control method overlaid with Kaizen concepts, I doubt most people other than me would engage in this kind of anality.  But it does work for me and it keeps me comfortable that I have a spending process that is both under control and tuned to an acceptable risk level for whatever age and environment I am in.   Note that this is where the income statement and the balance sheet meet.  

6. The Balance Sheet.  It seems difficult to comprehend a retiree who has not, if they have made any serious effort to successfully manage their retirement and/or develop their estate plan, created a balance sheet. This, along with the income statement is one of those sine qua non's of prudent self-management of retirement. Whole books have been written on this kind of thing.  I once worked closely with a family office and watched what they did. The very first thing they did was create a balance sheet to manage the client relationship (for at least some very HNW families this was a new concept!).  This document can be a very simple spreadsheet: Asset and Liability categories, item category and names, how titled, market value, cost basis, location, account number, monetizable or not for retirement, beneficiary desig, etc. etc. Some sheets will be more complex, of course, and other's less but keeping a list is helpful for financial and estate planning.  Remember, the success of your retirement will not be based on your brokerage accounts alone, it'll be based on your net worth and total balance sheet (that and your spending). Some asset allocation decisions need to know what is on the rest of the balance sheet to be optimal.  I realize many people have these but no small few do not. 

6. Feasibility Using the Actuarial Balance Sheet.  This balance sheet is the same as above but now we add the present value of "flow" and/or  future assets or liabilities to the sheet, things like the NPV of social security or pensions, or the NPV of future liabilities such as education costs, long term care, or the expected stream of future spending.  This exercise determines the feasibility of the retirement in present value terms.  Wade Pfau, I believe, calls this the funded ratio. I'd have to look it up but the basic idea is pv(A) > pv(L) is good, otherwise not so good. PatrickCollins, for his part, calls this kind of thing feasibility analysis. 

Since discount rates are often more art than science and since not everyone is facile with spreadsheets and discounted cash flow math (not as hard as it sounds) I'd highly recommend, if you are in that camp, that you outsource this task to the spreadsheets created by Ken Steiner at howmuchcanIaffordtospendinretirement.com.  Ken is an actuary and has made an entire post-retirement blogging career around helping people get their arms around this very important aspect of retirement planning. I consider his site best in class on this topic. If you can handle a spreadsheet and know PV math and aren't too rigid in discount rate theory, it's easy enough to do on one's own.  How it is used is another post altogether but an "unsustainable" balance sheet needs attention, probably by spending less or working longer.  At least you can see the problem. I don't think I can overstate the importance of the picture this creates to current and forward retirement planning.  

6. Sustainability Analysis Using Simulation.  Monte Carlo simulation has taken its fair share of shots lately for quite a few good reasons but it is still a useful part of planning since it is one of the few tools that can step outside of present value analysis and step outside of the constraints of historical data to show alternative possible scenarios and futures to which a retiree should pay attention. Not freak out, weep, and change everything...just pay attention and maybe feed that back into item # 3 above.  

One of the many problems with simulation is that the more complex it gets the more the underlying processes and the biases and assumptions of the designers (or even the back office staff stuffing in assumptions to the software interface) retreat towards either invisibility or incomprehensibility.  To solve this particular problem I built my own simulator (and I avoided bank professional service fees to boot ... I realize this is an option that is not available to everyone) but if I hadn't I would have used multiple models and done some kind of model averaging to diversify out some of this idiosyncratic risk.  For free tools I have used the following in the past though there are quite a few others if I were to look harder which I am not really inclined to do right now:

a. Firesim or cFiresim.  These tend towards the historical method of using rolling historical periods against the assumptions.  That's ok if you trust history which most years is reasonable. Maybe not this year but most years. At least one of these I think has added some sim capabilities

b. flexibleretirementplanner. This is a neat simulation tool and does what one might expect in a credible, trustworthy way within its own limits.

c. Other. I say other because I did not keep track but they are out there. Google "free retirement simulator."  Just pay attention to what they are doing and what they assume. In "other" I also include the planning software provided by advisors and brokers.  Some of these are actually really good.  I have no problem with good. The problem is that sometimes they charge you alot! and even if they don't you do not have access to when and how and how often they are run. Certainly you have no idea about the various assumptions that are embedded deep in the software.  Part of me wants to offer a service where anyone that gets charged more than $x for what is nothing more than a dressed up simulator, they can come to me and for $1 I will build a custom simulator for them.  Not sure that would work or if I could pull it off for more than about 5 people but it appeals to my subversive nature. 

d. My own custom simulator.  I did this more for fun (and to put up my middle finger to a banker that wanted to charge me 4K for the same thing) but I also can create infinite customization to test things out: variable longevity; weird spending in terms of spend trends, spend shocks, planned discontinuities, spend volatility, etc; different types and  durations of income streams; dynamic asset allocation; non-normal alternative return distributions; different shapes of mortality expectation; return suppression regimes; utility functions, fail magnitude metrics; etc etc. You get the idea.  The jury is out, way out, on whether this complexity helps anything in my planning. But it is free and customizable. 

e. "Simulation-like" tools that are close to but not quite simulation.  I'll list these under "other tools in the toolbox" below.  Examples might include things like what I list below in 7c, 7d and several of the equations in 7h.

How these tools are used, as I mentioned in item 6, is the million dollar question and I probably won't answer it in a satisfying way here.  For me, I think in terms of thresholds, trends, and triangulation.  Is there a difference between a 6% fail rate and a 9% fail rate? probably not.  Is there a difference between 10% and 90%? Uh, yeah.  So some threshold level makes sense but picking one is hard.  Academics have mentioned 30% somewhere in papers I have read but where does that come from?  It seems high even if it is meaningful.  Then, also, I think a trend approach works ok as well. This is akin to the statistical process control method I mentioned above for spending.  If ruin rate is trending down, cool.  If it it trending up, at whatever level, that might require a change in posture, from slouching in one's retirement seat to sitting forward on the chair.  Simulation can't predict the future but it can tell you when you need to pay attention.  Finally, since all models have strengths, weaknesses, and explicit or implicit biases, I like to triangulate across multiple models. This is not strictly a "model averaging" approach though that is ok too. Since some models are more sensitive than others to changes in assumptions they can act as an early warning indicator. The other less sensitive models keep you from over-reacting too soon.

7. Other Tools in My Tool Box.  Because I have been blogging on this topic for a couple years, I have collected a bunch of tools and equations that also inform what I try to do. I am not necessarily recommending these, nor are some of them really available, but since this post is more or less an inventory of "what I use myself" I felt compelled to call them out.  Here are some of my "other" tools:

a. Judgment. This may seem obvious but I keep my judgment aroused so that it is aware of what I am doing and so I don't go too far off the deep end. This may be the best tool of all…but with a proviso for the next item…

b. Behavioral awareness. I educate myself on destructive behavioral biases in investing, allocating, spending etc.  I am as prone as anyone to things like panic selling, anchoring, and confirmation bias. There are a million others.  Awareness keeps me from (sometimes) making dumb decisions. It also informs how I structure my setup.  An advisor friend keeps telling me I should just manage all of my own capital but this (behavioral problems) is one reason I like to get a little help which in the end isn't too expensive, though I know we could get it down a little more if I pressed. 

c. The Kolmogorov equation. This is a partial differential equation that evaluates the lifetime risk of ruin.  I have a spreadsheet form written in VBA I got from Professor M. Milevsky though I also re-wrote it in R just for fun and to see what it was doing.  This equation is akin to what one would get out of a simple MC sim with variable longevity.  It is thought of as an analytic approach vs simulation but if you look at the way it uses a finite differences approach to come up with a ruin estimate, it looks more like simulation than it doesn't though it really kinda isn’t simulation.  I find it useful because it is a fast summary statistic and the spreadsheet lends itself to really fast flexible what-if analysis.  One could argue that it is too simple (ignoring the fact that it is a PDE) and that all the unique things about a particular retiree are lost.  But then on the other hand we can argue, with a good hand I think, on the case for relative simplicity, you know, Bonini's paradox and all....

d. My "FRET" (flexible ruin estimation) tool for estimating lifetime risk of ruin. This models the same process as the Kolmogorov equation and comes up with the same result but without the differential equation.  It is transparent about what is being modeled and lends itself to simple analytic clarity (and speed).  While it uses a mini-sim to model portfolio longevity it is not really a Monte Carlo simulator -- depending on how you look at it.  That mini-sim feature, though, allows me to model non-gaussian distributions which I find useful.  I don't know yet of anyone but me that has or does this so this is not likely to be a tool for anyone but me and for me it is more of a novelty and a research tool. So far though I like it better than the previous tool.  

e. Perfect Withdrawal Rate tool.  I'm not even sure how to characterize this thing.  This math has been featured at earlyretiremetnow.com and by Javier Estrada in a series of papers on max withdrawal rates. There were some others by Clare et al., Suarez, and Blanchett, too.  The math calculates the withdrawal rate that, with perfect foresight (hindsight?) of a return series over a given period would result in a FV of zero.  When the return variable is animated as a random variable one then gets a distribution of withdrawal rates all of which would have been "perfect." I coded this process myself and then for fun added stochastic Gompertz longevity.  That means that after running this little tool, when looking at a particular quantile of "perfect" in the resulting distribution (let's say 5th percentile) the answer for "successful" spend rate compares well to my simulator, Kolmogorov, my JPA tool, Blanchett's simple formula, and some others.  That means that this is a tool that is playing the same game with the same underlying processes. But, in the end, this is a custom one-off tool so it's doubtful that anyone but me would have something like it and even if they did, how to use it is totally up in the air. I like it because it comes at the same retirement problem in a way I wasn't expecting. That's pretty cool.

f. Asset allocation Optimizer. I don't really use this much but I did build it.  Based on an article by Gordon Irlam I built a backward induction engine using stochastic dynamic programming earlier this year.  This is not an actively used tool but it does confirm prior biases I may have had about asset allocation and it shows me the way forward for different levels of wealth and time. This confirmation was doubly confirmed when I put the results of the optimization into a simulator that used the results to actively and dynamically change asset allocation in response to year and changes in sim-wealth. This exercise was a good tonic for the impulse to otherwise engage in downward sloping equity glidepaths.

g. Mean variance mapper.  I do not engage in much rigorous forward portfolio optimization though maybe I should.  But, since I have a moderate allocation to myself as an asset manager engaged in systematic alternative risk, I do use mean-variance math to look backward in history at the performance of what I personally do against random allocations of major asset classes as well as against benchmarks and my own passive investing.  

h. Collected Formulas.  I pick up formulas as I go.  I profiled them here on my blog. A reader's digest version of that page is:

1) Portfolio Longevity Using Milevsky's Fibonacci Formula
2) Spend Rate Using the Excel PMT Function
3) Blanchett's Two Simple Spending Formulas
4) Evan Inglis' Divide by 20 rule
5) Ken Steiner's Actuarial Equation (see section 6 above)
6) Gordon Irlam's Simple Rules of Thumb for investing and consumption
7) Milevsky and Robinson's Sustainable Spending Rate without Simulation
8) Estrada's Maximum Withdrawal Rate
9) Zwecher's Percent of Assets Committed to a Floor
10) my very own RiversHedge spending rule of thumb RH40
11) Milevsky's "All or or Nothing" optimal age for annuitization

i. Websites I go back to.  I use retirement websites like everyone else.  Here are some of the ones I tend to go back to more often than not. Most of these were in my 2017 RiversHedge awards post.

1) aacalc.com This is the best of the theoretically, mathematically correct retirement sites
2) TheRetirementCafe.com  Dirk Cotton
3) caniretireyet   Darrow Kirkpatrick
4) howmuchcanIspendinretirement.com Ken Steiner
5) blog.thinknewfound.com Corey Hoffstein
6) abnormalreturns.com Tadas Viskanta
8) awealthofcommonsense.com Ben Carlson
9) financialsamurai
10) Morgan Housel
11) alphaarchitect Wes Gray

j. [post script addition 10/26/17, I forgot about this one] "Geometric return with spending over multiple periods" strategy comparison tool.  This uses an inverted version of the math in "e" above.  The basic form of the equation is something I found at earlyretirement now in his tech appendix section.  This tool (custom built in R) allows me to compare two strategies with respect to the real effects of geometric compounding with volatility present.  I can also adapt it to reflect the presence of spending in the process, too. This is pretty neat because while I can easily estimate the effects of multiple periods and vol on arithmetic returns via an estimator at an infinite horizon (let's say one of several estimators is g = a - V/2 where g is the geo return, a is arithmetic return, and V is std dev squared) sometimes we retirees do not have infinite horizons and also that estimator ignores a spend constraint.  If I am 60 and I plan to whole-hog annuitize at 75 then comparing two strategies gets interesting over 15 years with spending where over infinity with no spending it is probably a no-brainer.  If I have a 7% linear return strategy with 25% vol and a 6% linear return with under 10% vol, at infinity it's no contest and strategy 2 wins (I think...the estimator says so anyway at infinity.) but at 5 or 10 or 15 years I have no idea and I especially have no idea when spending is layered in to this.  While this is necessarily an expected value thing (so who knows about one's own individual path) it is at least worth looking at what strategy might be better in the "near mid term." There is sometimes a "cross over" point that needs to be thought through. This is a trick I picked up in a paper by Richard Michaud.


So there you have it. Not exactly 14 pages of dreck but I was in danger of getting there again I think.  You'll notice I have not dwelt too much on last mile stuff (SS claiming strategies, optimization of which accounts to draw on first, and the like). That's because I'm still 59 and have a relatively low ratio of non-tax accounts. Also my own research has been focused on the underlying structures of retirement so my own tools reflect that bias a bit. If I was forced to sum up what I have been doing, and what I expect to be doing for a while, and what I really use, it is this:

- Have purpose and judgment to guide the way
- Use basic and obvious tools like income statements and balance sheets
- Be aware of destructive behaviors and use zen-like detachment along the path
- Triangulate amongst the many tools and approaches to get a more common sense answer
- Prepare to adapt every morning that you can still get up
- Embrace the math but temper it with what we started with: purpose and judgement

No comments:

Post a Comment