May 31, 2016

Some Random Off-the-MapThoughts On Retirement Spending

I've been thinking about the topic of retirement spending lately…but not about the obvious stuff.  For the obvious stuff, I feel like I have successfully worked through enough of what I believe to be a decent percentage of the "cannon" on retirement finance over the last five years. That cannon ranges from research by the academics over the last 20 years to advice, papers, and marketing materials by practitioners to various flavors of opinion -- and sometimes some pretty solid research -- by retirees that write or blog on the subject.  There is, it seems, no shortage of research and opinions on this topic (often, but not always, about "safe withdrawal rates").  There is also, unfortunately, no real crisp "answer" yet to the problem because it was, is, and will continue to be a fairly open ended non-solvable problem (whether the spending is fixed or dynamic, "safe" or otherwise) because it depends on way too many other things. 


But that "mainstream obvious stuff" is neither here nor there.  What I have noticed, and what I want to post about, is that I feel like there seems to be a little bit of a divergence between what I see in the literature of retirement finance on spending and withdrawal rates and the real experience of it in real life by real retirees. 
Most of the writing I see leans towards something that is math-y,  unhelpfully abstracted, and maybe a little prescriptive.  That's why when push comes to shove I tend to trust retiree writing on the topic…all else being equal.  The academics and practitioners are either paid for this kind of thing or are at least still getting a paycheck.  The retirees (and their families one would presume), on the other hand, have skin in the game and feel the spending issue as a real, human, tangible phenomenon.  This is also why I wanted to disgorge out of my head what I might call a little mini-phenomenology of retirement spending as I have seen it so far.  So, for a minute or two let's forget about a direct, mainstream discussion of things like withdrawal rates, the 4% rule, or tricky "dynamic decision rules." Rather, let's walk through a list of 13 things I have observed on what I think are some of the under-appreciated aspects of the retirement spending "problem" especially as it relates to how one might personally experience it -- and that's with or without the math as the case may be...

1. "Safe Withdrawal Rates" have almost nothing to do with retiree spending as experienced in real life as I see it. 

I have to concur with Ken Steiner at HowMuchCanIAffordToSpendInRetirement.com in a post he did a while back on "Calling Off The Search For SafeWithdrawal Rates."  Among his many solid points were these four that I wanted to point out:

  • It [SWR research] ignores certain types of expenses such as long-term care expenses, other unexpected expenses and bequest motives. 
  •  It is designed to “draw-down” accumulated savings; not be part of a “bigger picture” spending budget  
  • It assumes that each year’s retiree spending will exactly equal the safe withdrawal amount. 
  • It contains no adjustment mechanisms to keep future spending on track if investments fail to earn rates assumed in the model (or investments earn more than assumed) or if actual spending deviates from the safe withdrawal amounts.
The main thing for me here is the third point on withdrawal: I don't "withdraw," I spend. I may have a goal of drawing down or withdrawing x% of a portfolio (say 40k on $1M i.e., 4%) but if I happen to spend only 30k I don't spend more to true it up to 40k (if I have "withdrawn" 40 and spend 30 then 10 gets reinvested or is now considered part of the cash allocation) nor do I somehow magically unspend 10k if I happen to spend 50k (though I might borrow from myself to smooth things out).  My spending varies as it varies…within a planning range…I hope. But it's always permanent; we always relentlessly consume.  That means spending is the only thing and I never really withdraw as such. Withdrawal, to me, is an academic's term (unless we are talking about the strategy of tapping different types of accounts at different times, something I do not touch on here).   

  
2. Spending is one of the very few high impact things one can control in retirement.

Forget fixed or dynamic withdrawal rules and success or fail rates for a second, let's focus back on what I already know is very self evident: spending (or withdrawal) is not just part of an abstract formula for retirement success, it is among the very, very few things over which we have any control and that also has any significant leverage on enhancing or correcting a high risk retirement situation (other than Social Security or a Plan B job, I suppose).  For example, here is my personal list of the things that I think a retiree can and can't control:

What one can control in the retirement equation
            - Starting endowment (up to a certain point)
            - Asset allocation (muted impact within a broad range)
            - Spending 
            - Lifestyle expectations
            - Health (i.e., longevity to a certain extent)
            - Legacy goals
            - Outside or plan-B income
            - General risk tolerance
            - Willingness to trade off present$ vs. future$
            - Volatility tolerance: spending or returns
            - Personal taxes on the margin
            - Soc. Security claim dates
            - Probably some others I'm not thinking about
           
What one can't control
            - Market Returns and volatility (except via portfolio design)
            - Normal spending volatility, to a certain extent
            - Longevity and age, past a certain point
            - Inflation
            - The sequence and volatility of returns
            - Spending shocks
            - Tax policy
            - Soc. Security policy or solvency
            - The future
            - Retirement start, past a certain point (e.g., health issues, layoffs, caregiving)
            - Probably some others I'm not thinking about

Of these various items, assuming the endowment value is already final, only spending seems to me like the really big main control lever (along with lifestyle expectations and risk tolerance, perhaps) -- with the proviso that control, in the end, does not always mean total control. More on that later.  This general point on spending is so obvious that I'm not sure why I even include it here other than to emphasize and re-emphasize to myself its singularly unique role in the retirement equation. It stands out. 


3. Direct spending is only one of many things that conspire to consume the nest egg and reduce retirement success.

Don't forget hidden or shadow spending.  This idea includes but is probably not limited to: taxes of all kinds, government fees not officially called taxes, advisory fees or commissions, fund fees and expenses, depreciation on autos or other depreciating assets, deferred home maintenance - a cost that relentlessly accrues whether spent now or not, the unplanned marginal health care or education costs that outstrip inflation, excess cash balances that under-perform inflation, the unspent but silently accruing future support obligations for adult-ish children to whom it will be difficult to say no, fraud and identity theft, maybe an accumulating risk of divorce, chaotic spending shocks, etc. All of this stuff consumes an asset base as surely as normal spending does.  Some of this is naturally included in a family income statement, some not. Sometimes these show up in retirement calculators and articles, sometimes not. There are probably many other examples. Real spending over the long run is generally more challenging than it looks in a simple model or on paper.


4.  Retirement spending is not necessarily "one number" or a constant (or even always predictably and controllably dynamic), it is by its nature a random/stochastic process.

So many things that I read in the retirement finance zone make it sound like spending is somehow like some precise little mini-monolith.   You pick a spend rate and that is "the number."  Maybe one does some adjustments or uses some slick dynamic rules or something but the sense is that spending is somehow totally deterministic and completely under one's control: one number, adjustable perhaps, but one number. The reality is, of course, otherwise.   Here for example is a comment I found on Slate a while back about spending in general that could and should easily apply to retirement spending:

            As it turns out, the primary issue may simply be that the concept of a steady monthly financial budget does not conform to the realities of most people's financial lives, which are not nearly so consistent. A recent report from JPMorgan Chase Institute found that more than 80% of the bank's customers experience a greater-than-5% change in both money coming in and their overall spending on a month-to-month basis; 25% of us see a variation greater than 30%(!) from one year to the next! This perhaps isn't surprising, given that income itself fluctuates more in our increasingly freelancer-based economy (from outright freelancing gigs, to those who earn money through companies like Uber or Airbnb). Though notably, the study also found that spending changes don't necessarily align neatly to income fluctuations; 60% of us vary our spending by more than 30% at least occasionally, even if income hasn't changed, given all sorts of moderate "spending shocks" that can arise (from outright emergencies to pet emergencies to just the irregularity of paying for kids' summer camp or gifts during the holiday season).

So, statistically speaking then, spending is random -- around some center or average or mode, perhaps, but random.  There probably are even two types of variance, if we think about it carefully and put it into a statistical process control context: little variance or random noise and big variance or chaotic big shocks.  Generally one can influence the first and has a harder time with the second, but more on that later.  In addition to what I consider the easy-to-understand first-order random variance (normally distributed or otherwise) there are no doubt second order types of spending variance, too.  For example: a) the spend variance can expand and contract; instead of +/- 20% this year maybe it's +/-30% next year…or maybe it's +/-10% instead, and b) the center can drift randomly or trend; you planned on a 4% spend -- perhaps even with some variance -- but the real spend average drifts up and down or maybe just trends up…which is my personal experience.  Now try to tell me with exact precision what you think you will spend next month or next year.  You can't do it. Maybe this kind of thing won't have a material impact on success rates in standard retirement software but it does at a minimum screw around with planning. 

5. Spending randomness is more than likely not normally distributed and that skewed randomness can perhaps have a negative influence on how long the portfolio lasts as well as possibly undermining spending stability over time. 

Intuition and personal experience (I attempted to barbarically and ruthlessly cut my spending by half five years ago) made me hypothesize that random spending variance is not normally distributed.  It is much harder to curtail spending than it is to expand it. And in fact, when I looked at my monthly spend variation over the last 5 years or so it looked like this. The average spending variance is pretty close to zero but it clearly is random and it clearly skews north:


       

So, before we introduce any variance, if one person were to hold everything exactly the same (no returns, no inflation, no spending volatility; yes I know that's dumb) and he or she spent exactly 40k per year, the portfolio lasts about 25 years.  On the other hand, if one were to do the exact same thing but then let spending vary in a skewed way like the above chart [for the next chart I made a fake distribution similar to the above except wider with an average of 0 and a slight skew up in overspending] and then simulate it 10,000 times, the end portfolio variance (compared to the previous 0 value) at ~25years would look like this: 



The end game varies!  No big surprise there, but first a few comments on this exercise: 

First of all, life is not a simulation, you only get one bite at the apple, not 10,000 simulations. But that maybe makes this analysis even worse because it means that your one bite might be the one run at the far left; and that's about a year less of solvency in this particular model.  

Second - this particular example of variation would probably be totally swamped and made totally invisible by the real variation effects that happen with random returns and inflation thrown in. So maybe this little tiny point is mooted by all that bigger stuff. I'm just trying to make a general point about random spending variance.

Third - the result above does skew left a bit but it's hard to see.

Fourth - you don’t really have to simulate this stuff -- which is kind of an overkill type of analysis -- for making the point; I just happened to have a tool at hand when I was thinking about it.

Sixth - if you were to also let returns and inflation vary as well as spending, you would in effect -- by accidentally/randomly letting spending go high when returns are low and/or inflation high -- have another small contribution to "sequence of returns risk."  By that I mean that if spending a fixed 4% is bad when the portfolio goes down for a while early in retirement, its probably even worse if you tend to get a random throw higher in spending at the same time that the portfolio goes down.

Seventh - while I assumed an average value of zero for the spending variance distribution, my personal guess -- based on nothing at all -- is that an up-skew in spending in real life would have a slow gravitational pull on average spending i.e., I think the spending "average" would be dragged up over time by nothing other than the control-less skew…in the absence of spending controls that is.

Eighth - using monthly data makes it look worse than it probably is. Some of this gets smoothed out when looking at annual data though it doesn't go away.  Anyway, I work on a monthly cycle so that's what I think about.

6. The skewed randomness in spending can add a quite a little bit of fun and drama to an otherwise normal "dynamic spending" scenario.

John Walton, an engineering professor and soon-to-be-retiree, in a great article on the effects of dynamic spending in retirement at AdvisorPerspectives, lays out his idea on the full spectrum of retirement spending/consumption within which most of the modern spending rules in the current RetFin literature seem to fall.  Starting with the 4% rule at one end of the scale he points out -- like others have and like I did here on how spending would have felt in 2009 -- that the 4% rule in effect changes all the time based on your current portfolio value(P). If P goes up, effective spending (if it is constant) is lower as a % of P and if P goes down spending goes up as a percentage.  He calls that a negative tilt, a tilt that may help conserve lifestyle by maintaining a constant spend but also comes at the cost of lower success rates and possibly higher sequence of returns risk.  Going up the scale, he calls a constant percentage spend of P (the endowment model) a zero tilt.  This is where the portfolio will never fail but your lifestyle might take a beating now and then.  Going even further up the scale he describes a "positive tilt" where when P is up one spends more and when P is down one spends less, accurately and helpfully pointing out that in practice this is how most -- or at least many -- retirees behave in real life.  The main thrust of his article then, if I have it right, is that: 1) this spectrum is more or less the range boundary of many of our current spending decision (or withdrawal) rules,  2) negative tilts including a fixed spend like the 4% rule can cushion one's lifestyle but may book the sequence of returns risk, and 3) positive tilts, for the same reasons, can capture a long term sequence of returns "bonus," provide higher long term success rates, but maybe mess around aggressively with lifestyle.

Whew.  Let's take a breath here.  Ok, so I think this is a pretty good way to contextualize retirement spending and I think it clarifies if not adds helpfully to the literature on RetFin.  But wait! It gets more interesting than he lets on if an OCD retiree like me were to look too closely and too obsessively at all of this.  A path not taken in his discussion, though it is implicit, is that one can go quite a bit to the "left" of the 4% rule on his spectrum into much darker negative tilt territory.  4% is not the dead-end terminus of the spectrum, it is merely a way station.  What I mean by that is that, hypothetically speaking, if a person were to: a) engage in willfully destructive high spending when a portfolio is down, b) in addition be exposed to a strong factor of random but unlucky skewed high spending on top of that (hey, I know streak theory, hot or cold, has been more or less debunked but also remember the "one bite at the apple" theory I mentioned earlier), and c) have a much harder time reducing spending than increasing it (the source of the skew), then the path seems wide open to the left of the 4% rule on Walton's "tilt spectrum." That zone to the left would represent a type of power-version of sequence risk…and I haven't even touched on spending shocks yet.  That kind of spending setup looks like it would be pretty painful for long term retirement success rates and it would probably feel, if one could actually sense it at the time, like a free fall -- if the lurch to the left were big enough and long enough over the wrong time frame.  Fail rates, by the way, on any tool you want to pick would look really, really high.

But honestly, who would do that kind of thing to themselves? Oh yeah, that's right, I would…for a year or so on either side of my retirement.  It happens.   The only thing that has softened the risk for me, in my opinion, was a pretty assertive reach, in the 2nd year after retirement, for knowledge on how retirement finance works, a ruthless cost reduction, and the beneficence, if not providence, of a world class bull market since 2009.  Would that we all had that kind of thing when we needed it.  Me? I got some gratitude…


7. There is more than one type of destructively high spending…or why dynamic decision rules can sometimes be a chimera 

In item 4 above I described how spending is not necessarily one choice or one thing, it has a random component that makes it hard to manage.  In item 5 I described how the randomness can skew in a bad direction.  In addition to those two points I have to add that many -- but not all, of course -- retirement finance articles I read make it sound like spending is perfectly tractable and that adaptation is easy and decision rules are simple to execute. All true I suppose, except that there is always more than one reason for sustained high spending in the face of a risky retirement. For example:
           
            Spending might just be flat-out high as part of a personal policy, in full awareness of the probability of failure (the assumption might be that Soc Sec will save me or an inheritance is coming or some future job is in the bag, or maybe just arrogance despite the math, etc)

            Spending might be high but there might be a total unawareness that spending is high and/or there is no idea of the key elements of the various retirement finance equations or maybe there is not a compelling sense of risk or maybe risk aversion is unusually low...

            Spending is high (or medium or low for that matter) but there is no control or limited control on spending changes.  For example, one might have a spendthrift partner.

            As mentioned in item 5, spending changes are more likely than not hard and intractable on the way down and easy on the way up.  The degree of up/down difference may be different for different people.

            Circumstantial short term spending might take a budget way out of range for a finite period of time e.g., education costs or temporary health costs, care-giving and support of others, etc. Some of this can be budgeted and predicted, some not. These semi-predictable expenses might not quite fall into spending shock territory.

            Big, debilitating, sometimes bankrupting spending shocks can occur that are less a random process than a chaotic one (see Dirk Cotton on retirement income and chaos theory here along with follow-on posts on the material impact and causes of retirement bankruptcy and the cascading effects of multiple problems that interact at the same time…divorce, health shocks, identity fraud…here and here)


Some of the items in this discussion point are pretty self evident and the ones that maybe aren't are covered better elsewhere (e.g., theretirementcafe.com). So why do I reiterate this stuff here? Like item 2 above, it is to remind myself, more than you, that this stuff happens. In addition, I should mention that I've also found in my personal experience that aggressively low spending in retirement can sometimes be called out by others as unnecessary self-denial (sometimes this is done correctly and helpfully but sometime it is done insultingly by, for example, a now ex-advisor).  Ignore them.  It is sometimes clearly a wise and necessary accrual for the future liabilities that the universe can throw down to try to break us. 


8.  I think that academics or practitioners that recommend spend rates like 5 or 6% are irresponsible.

I have read more than a couple articles over the last year or two where the author makes a claim, based on backtesting or whatever, that via some deft use of decision rules and/or through a rosy interpretation of history or Monte Carlo simulations that a retiree can spend more than 4% without worry, sometimes into the 5% or 6% range. I think this is totally irresponsible.  This is because:

            a. They have no skin in the game and you, the retiree, cannot un-ring the bell of a bad series of retirement years.  They can run 10,000 simulations and give an opinion; you can run but one life.  Where will they be when you run out of money, social security is a pale ghost, and your kids have abandoned you.

            b. Yes I get the whole risk of self-denial and over-contributing-to-legacy thing but rewind to my previous item above on the possibility for the in-adaptability of spending and then contemplate this question again.

            c. The article writers are either unaware of or don’t distinguish between the concepts of "two retirements." This is the idea that the mathematical and experiential differences between a traditional retiree and an early retiree are sufficiently unique that they can be considered two different things.  Darrow Kirkpatrick in his new book "Can I Retire Yet" does a great job of laying this out (p. 13). That means that I think that "early" retirees who happen to choose to spend 6% in retirement are cliff diving into a sink hole.  Those that recommend it are engaged in a type of con. 

            d. Market history -- and what seems like analysis by everyone and his brother in the econoblogoshpere -- makes it sound like future return assumptions look pretty bleak for almost everything as of Q1 2016.  Maybe you can spend 6% into the face of that storm but I won't if I can help it.  Maybe it'll all work out but that would be a bonus.      

9. I've heard that there are actually people that think tracking spending (and income) is too much of a hassle.

If none of the items above or below can convince you to carefully track and think about your spending, you are probably beyond help.  This is from Darrow Kirkpatrick's new book on retirement:
"The starting point, the most critical factor and yet the most neglected one, when planning for retirement, is your regular living expenses. Without a deep understanding of what it costs you to live, any discussion of retirement savings or income is pointless…Determining expenses, because it requires discipline and detail, is hard for many people."  [emphasis added]
This is a post about spending.  To think about or read about or talk about spending without then knowing and tracking spending would be an absurdity of the first order.


10. The impact of spending on risk is fairly well known but I don't think it is really truly "humanly" appreciated and maybe it's even a little mis-analysed sometimes. 

A. Ruin rates accelerate above 2%.   Back in item 8 I mentioned that articles by practitioners sometimes hint at a 5-6% spend rate and are, in my opinion, irresponsible (Wade Pfau goes towards 2% in some cases but he is an exception).  Here is another reason why I think this way.  Take any calculator/model you wish and factor in ever higher spending rates.  The risk, when evaluated over 30 years of retirement (I'm not even thinking about the 40-50 years an early retiree could see…yet), starts to accelerate after about a 2% spend rate. The risk above 4% is very non-trivial.  I didn't run the following chart with the benefit of spending rules [should I test that?] but one might see an acceleration there too.  Here is the fail rate estimate of fixed spending at different levels using five different tools:


There are some differences here of course (the underlying assumptions don't perfectly cohere but they are close enough) but the general consensus seems to be: watch out above a 3% spend rate.  Some researchers like to suggest that we should avoid discussing ruin probabilities altogether but they also say to maybe take a warning after a 30% fail estimate.  That seems about right.  Darrow Kirkpatrick in his new book, based on a variety of sources, suggests that 3-5% is the range and 3-4% is maybe better.  That sounds about right, too.  A recent run I did with the variable spending model implied by ARVA (Waring and Seigel's Annually Recalculated Virtual Annuity) led me to believe, based on forward expectations of inflation and returns, that 3-5% was probably a good range for spending with under 4% feeling a bit better than above.  That seems to be in the same ball park as the rest.  Of course there are the original Bengen and Trinity studies but it looks like they are in same game, too.  Smart decision rules might get you over 5% but in that you are on your own, In the end, I don't really think that this particular risk of accelerating fail rates above 2-3% is either unknown or un-analysed by others, I just think it is under-appreciated.

B. The 1970s are my touchstone for safe vs. unsafe retirement spending.  I survived 2009 and I also wrote a couple posts recently on how spending might have felt for a retiree with a fixed 4% or 3% spend rate in '09 (here and here). The 1930's are an abstraction that I have not approached yet so I reserve my main fear for the 1970s.  And that's not necessarily because of disco or leisure suits although that would be enough right there.  And it's not necessarily because of a combination of high inflation and low returns, though that will really be the main point at some point.  The 1970s conjure fear because I was intimately connected with some of the downside of the 70s.  I went through high school and college and my first job search back then. In addition, I was both the dependent of and, eventually, much later, a caregiver for a parent that lived on a small private portfolio during much of that time. I saw first hand what high inflation and low returns did to people during those years.  2009, for all its drama, has nothing on a decade or more of stagflation (and we have had the luxury of an historic bull market since '09). 

Bengen and the Trinity Study supposedly used assumptions for retirement spending and portfolios that were designed to survive the very worst periods like the 30s and 1966-1982. That’s where the 4% rule came from I hear and also why it is supposed to be robust and why layering on more complex decision rules is supposed to create the opportunity to optimize spending to a level higher than 4%.  I came to a slightly different conclusion recently.  It would be awfully hubristic of me to say I was smarter or a better modeler than the big names of the past, neither of which I think is true. Let's just say I used different assumptions and even if I'm wrong, which I might be, at least I'm more conservative, which can't be all bad.  First, here is a chart of returns and inflation over that period (70s) using a rolling 12 month kind of thing (did you know there was such a big drawdown in 1974? I didn't. I was a sophomore in high school):



This is a pretty ugly chart but that is what I was looking for to test risk.  Given the ugliness, I tried to model a $1M portfolio thru this period (for a 57year old, a modified 60/40 portfolio and a fixed spend rate) to see what happens.  I had some differences from the prior studies.  The main differences that I can think of between me and the big names might be that I:  1) used different data sources, 2) made big mistakes, 3) used a total return calc* for bonds (5year treasury) vs. yield returns, 4) had a zero cash assumption with reallocation, 5) I threw in a little tiny allocation of gold, 6) a ton of other reasons… In any case, I happened to come up with a big fail for a fixed inflated 4% spend but I also came up with a 3% spend that was "just OK."   Here, if I was correct, is what a 4% and 3% fixed spend looked like as an "effective" real time rate as the portfolio varies. Ignore the ARVA line for now.



Ok, let's make a bold assumption that I got this right.  That 4% spend is a really big fail unless one were to keel over before age 83.  3% seems to work over the long run but you have to zoom into that rate to tease out  my core point.  Even if 3% worked through the 70s and early 80s, it was experientially (as far as I can tell from the chart and my memory of my mother's situation) quite difficult to live through.  Here for example is a close-up of the 3% spend model run from Jan 1970 to around 2006 with the forward fail rate estimates superimposed.  While looking at this keep in mind that a lot of commentators say that a >5% spend rate and a >30% fail rate are the threshold points for "let's start to worry."  




What do I see? I see that a 3% starting spend rate more or less works but I also see that after about 9 or 10 years the effective spend rate was greater than 5% for most of the time thereafter -- let's call it 23-26 years on this chart. That's a long time. I also see that the forward fail rate estimates might have been greater than 30% for at least 8 years or so.  I don't know about you but this is not a recipe for a relaxing retirement. It is, however, a recipe for emotional and unconstructive actions i.e., bad behavioral finance.  4%, on the other hand, would have been a terrifying uncontrolled free fall. 

If I got all of this wrong, you, my friend, are free to go on your way and spend as much as you wish whenever you wish.  Me? I am still intimidated by the possible onslaught of bad returns and inflation at the same time and I will now always view 4% with suspicion and 5%+ more as, what did I call it? Irresponsible. My ex-broker notwithstanding.  
"Unpredictable volatility makes investors more susceptible to the “freakout factor”—when they suffer a loss big enough in their portfolio that it triggers an emotional, and often unproductive, move."
(Andrew Lo)
C. The concept of "two retirements" adds a layer of misunderstanding on top of the normal risk of spend rates. As I implied somewhere above, the old, traditional models for retirement generally look at 65 year olds with an average life expectancy of, let's say, 20 years or maybe 30 years at the most (if you were to be 70 and smoke and have heart disease and pile of money, you are, from a mathematical point of view anyway, in pretty good shape). On the other hand, for an healthy early retiree with  40 - 50 years of potential longevity and "just enough" capital, you are totally on the razor's edge of retirement risk.  It is a completely different retirement than the traditional model.  I think that this risk is quite a bit underappreciated because "early retirement" is more or less a modern phenomenon and to my thinking a little under-researched.  Some good work is going into this area but mostly it's from retirees rather than the academics or practitioners who seem to focus more on the traditional model. 

Here, for example, is a chart I ran a few years ago that uses a couple popular free retirement calculators, among others, to see what  happens if the duration of a retirement is pushed out for a fixed 4% inflation adjusted spend.  I don't know if this is the right way to model this but at least the different tools look more or less consistent: Fail rates accelerate after ~25 years.  I'd be curious to see what a dynamic spend policy can do here but the take away is that I would not want to hang on to a fixed 4% spend for a retirement lasting 40 or more years.  The 3.5% spend rate that I ran through an historical sim tool (dotted line) looks better but even there it looks like one would have to keep a close eye on spending for really long retirements.  


We're a long way from the header of this particular item of mine but let's repeat it to close out here:  "The impact of spending on risk is fairly well known but I don't think it is really truly "humanly" appreciated and maybe it's even a little mis-analysed sometimes."


 12. There are a ton of good reasons to try to control spending.

#12 is not really a new point. It more or less reiterates much of the above: spending is both a core lever and a core risk of a long retirement…so some basic points are: 

            a. The random variance of spending can affect portfolio outcomes over time
            b. Random skewed spending variance and willful bad behavior can take spending completely "out of phase" with what really works over the long run.
            c. We don't simulate our lives, we have one bite at the apple. We  better get it right the one time we have.
            d. We are often insufficiently aware of the potential for random or run-away or chaotic spending to sabotage retirement and we probably need to watch it and control it better.
            e. (Read the literature of retirement for all the other main structural points on spending risk. This post doesn't cover the main good stuff)
             
all of which is a lead-in to...

13. To my eye, retirement spending, given all the risks and opportunities described above, looks like it could be like an industrial statistical control process and could also benefit from a little "Kaizen."

Points 1-12 makes me want to believe point 13 will help.  I hope it will. Maybe its just retirement OCD.  

Statistical process control, what a team and I used to do with software development back in the 90s, in my mind is like this: 1) be aware that a process exists in the first place, 2) start to measure it, 3) create a  baseline, 4) seek to improve the process through data analysis and improvement methods, 5) measure again, improve again, measure again, etc. When it is stable and demonstrably improved, consider optimizing the process one way or another.  Here is Wikipedia: 

            Statistical process control (SPC) is a method of quality control which uses statistical methods. SPC is applied in order to monitor and control a process. Monitoring and controlling the process ensures that it operates at its full potential. At its full potential, the process can make as much conforming product as possible with a minimum (if not an elimination) of waste (rework or scrap). SPC can be applied to any process where the "conforming product" (product meeting specifications) output can be measured. Key tools used in SPC include control charts; a focus on continuous improvement; and the design of experiments. An example of a process where SPC is applied is manufacturing lines.

And, for what it's worth, let's talk about Kaizen.  This is an idea tied to Japan but originates out of early 20th century American industrial experience if not before. It is usually but not always tied at the hip to SPC. W. Edwards Deming gets the main credit for this kind of thing but there are other players.  Underlying Kaizen (or continuous improvement or 6-sigma) is the basic principle of an iterative Deming-style improvement cycle that looks like this: Plan, do, inspect, act(fix), repeat.  Here is Wiki again: 

            Kaizen, Japanese for "improvement." When used in the business sense and applied to the workplace, kaizen refers to activities that continuously improve all functions and involve all employees from the CEO to the assembly line workers. It also applies to processes, such as purchasing and logistics, that cross organizational boundaries into the supply chain. It has been applied in healthcare, psychotherapy, life-coaching, government, banking, and other industries.  By improving standardized activities and processes, kaizen aims to eliminate waste.

One of the major, but of course not the only, tools of SPC is the process control chart.  It is used to visualize the process and gauge the extent of and consistency of improvements.  It is this kind of chart/process that I think can apply directly to retirement spending.  One can Google (images) "process control chart" and find a ton examples as applied in a variety of industries. Here is one example for a "flange manufacturing" process:



This shows the basic steps pretty well:

- measure a baseline and determine the mean and standard deviation
- try to improve the mean and std deviation by whatever means
- measure again and show the results
- try to improve (or optimize) the mean and standard deviation again
- measure again and show it again and then do it iteratively

This, frankly, feels like a good fit with the idea of retirement spending process control and improvement (if one were to believe and fear any of the above points on spending risk).  Here is what a "spending control process" might look like if it were rendered in terms of spending as a percent of a portfolio.  This would, however, require one to rise above item 9 above i.e., the reluctance to look closely at and measure expenses. 




red - Monthly spend as a % of portfolio
blue - 12 month moving average of the red line
yellow - 2standard dev (12 month) of the red line +/-the blue line
red dotted - upper and lower control limits
grey dotted - mean expected value for time frame

Stage one is the unsustainable, out of control process, i.e., the baseline.
Stage two is "be aware, analyze, take action, improve, control"
State three is: build on stage two and figure out where to go next...optimize

Ignore for now that the 2009+ bull market might've helped out the spending percent on this chart (also, ignore how the denominator is exactly calculated, you can come up with your own) , focus on: 1) the out of control unsustainability of stage 1, and 2) the material mean and standard deviation changes seen in going from stage 1 to stage 2. 

This is where it gets interesting, though.  Once the process is visible (some contend that the mere fact of visibility and measurement will drive down spending) and under control, the next step of planning and optimizing can be done on a coherent basis.   This is also where I think the concept of the upper control limit (UCL) visualized as "the X% (e.g., 4%) rule" finally comes in.  The UCL, a standard in statistical process control methods -- the upper bound of a process under firm control -- is probably, in my interpretation, what someone else might call a safe withdrawal rate.  That means that whatever math you use to determine what you can spend, maybe conceive of that spend as an UCL, then make sure your process is under reasonable, narrow control with the UCL as your upper boundary. Maybe use the average but I like UCL.  If your are an early retiree, your boundary should probably be pretty conservative, by the way. Practically speaking, I think that almost all of point #13 is a little anal , perhaps, and probably tests the boundaries of "too much effort" for some but then again, as an economist might say: running out of money has infinite disutility.  How much effort would you take to avoid that disutility.    


That's it.  That is my mini-phenomenology of spending.  In the end I guess my main point is that retirement spending is not really math even though it looks like it is. It is lived experience. It is the source of much retirement anxiety as well as a really big lever on making it all work. I'd take it seriously. 


----------------------------


* I built total returns on a variation of the following formula which I confirmed with  Aswath Damodaran at the Stern Shool, Wade Pfau, and the developer behind cFireSim.com.  The formula was adapted depending on whether I was using 10 year or 5 year data. Data was from St Louis Fed data series.
bonds(i)=100*(yields(i)./yields(i+1) + (1-yields(i)./yields(i+1))./((1+yields(i+1)/100).^10) - 1 + yields(i)/100);





No comments:

Post a Comment