Spending Declines in Retirement?

After reading a recent post by Dirk Cotton I wanted to take a look at spending reductions in retirement as it relates to a naïve and simplistic assumption I came up with, in it's first form, more than 10 years ago to model lower spending as I age (created before I had read a lick of retirement finance and that has followed me for better or worse in my planning ever since).  Then I wanted to compare that personal assumption to a couple of new-ish research-based ways of looking at the same thing (empirically demonstrated spending declines at later ages) that were mentioned by Cotton in his post. 

My naïve assumption at the time[1] was that I might be able to spend in three stages: A) spend at a relatively high level now while I still have young school-age children, B) spend at a slightly lower level when my youngest gets out of high school and then all the way to age 85 or what is close to the median expected longevity for me, and C) after age 85 budget for some minimum level of spending to age whatever (kind of a self-rolled-annuity in a way, minus the risk pooling) on the assumption that I have a better than 50% chance of not being around anyway.  This three stage assumption allowed me to not over-estimate the actuarial value of lifetime consumption for my balance sheet.  That meant I could shift, at least conceptually, some spending from the future to today and not reserve so much on the balance sheet for an uncertain and unexamined future.   When I looked at the data (e.g., the last 12-24 months of spending) and then dug in to the assumptions (i.e., what, exactly, on the income statement would go away when youngest leaves the nest…and what, exactly, might I need in today's dollars to survive at some minimal level past 85, ignoring things like health crises and LTC?) I came up with this spending budget idea for retirement: A) maintain status quo spending to age 68 (and just for fun let's say that that status quo is after shaking myself down for a 50% reduction in spending), B) plan for an approximate 20% decline at age 68, and then C) budget for an approximate 42% decline at 85.  The purpose of stage C was to budget for a quasi-survival level of spending that would be higher than living under a bridge but lower than living the high life at an age when I probably won't even be there.  Since we are talking about spending/consumption here and not income we are ignoring sources of income like Social Security.

Against all of this stuff (in other words, my naïve, simple, and personal assumption above) I wanted to counterpose at least two or three research based ways of looking at spending declines that have been articulated in the time since I came up with my "rule" (e.g., Blanchett 2013, Banerjee2012, and Cotton summarizing the previous two 2016) and then see how close I got to what research seems to be telling us about retirement spending. To oversimplify all of this -- and perhaps misapply the research while I'm at it -- this is what the various spending assumptions look like to me:

Me: a simplistic step function: status quo spend to age 68, -20% at 69 (-19% but why quibble) and -42% (~29% incremental) at age 86+.

Blanchett: a regression formula in the form ΔAS = .00008(age^2) - (.0125*age) - .0066ln(ExpTar) + 54.6% where ΔAS = change in real annual spending; ExpTar = after tax total expenditure target.  Read the article for where this comes from.  To quote David B: "…using the spending curves based on actual retiree expenditures, we see that the total need decreases…throughout retirement."

Banerjee (distilling his own research down to the basics): start the model at 65, reduce spending 19% by age 75, 34% by age 85, and 52% by age 95.  Note that in the graph below I do a linear interpolation between those points.  

Dirk Cotton summarizing the previous two: plan for spending declines of between 1.5 and 2.0% over a full retirement. Let's assume that starts at 65. To quote Dirk: "We shouldn't plan on spending the same amount when we're 80 as we do at 65, though David Blanchett and Sudipto Banerjee have shown that spending typically declines roughly 1.5% to 2% a year as we age." 

When plotted out by age, this is what it all looks like to me. Ignore for the moment whether I mis-applied the Blanchet Formula or whether I have properly understood how the authors factored in inflation and "real" spending changes.  Also assume for the moment that at a minimum I'll plan on discount rates = inflation so that I might be closer to getting it right than not.  Also, fwiw, I started the rules and formulas (except mine) at age 65 and ended them at 95 because I think that was how they were thinking about it…more or less:

Before anyone[2] starts to over-analyze this kind of thing, let's be clear on what this analysis is NOT:

• Can I or anyone predict future spending (or future anything)?
• Are spending declines even real, research notwithstanding?
• Am I exactly right in my modeling?
• Are Blanchett or Banerjee right?
• Did I apply Blanchett's formula correctly (maybe not!)?
• Do I think spending might actually increase rather than decrease for a while?
• Are Monte Carlo sims a good thing or a bad thing (implied in Blanchett's work)?
• Are static or dynamic spending models a good thing or bad thing?
• What about Wade Pfau's research?
• Are there better models or formulas elsewhere?

On the other hand, this is what the intent of my analysis really IS for better or worse:

• Did my original -- and rather blunt and dated and simple and unexamined -- intuition on spending get even close to a couple of pros doing some research that I hadn't seen before?
• If I continue to use my own custom MC simulator should I now try to make it even more sophisticated on spending declines to reflect some of this research? I already have my own rudimentary assumptions baked in.
• Is this declining spending approach even constructive for me at this point from a planning point of view?  I have been using something similar in a naïve way, but does it or should it mean anything? 

So, any conclusions from all of this?  Sure, but let's look at "conclusions" in terms of the previous three bullet points:

1.  Did my original -- and rather blunt and dated and simple and unexamined -- intuition on spending get even close to a couple of pros doing some research that I hadn't seen before?

Yeah, I think so.  It looks to my untrained eye like I got it about right...except for the thought that the step function is maybe a little too broad over longish age ranges.  In general, though, it seems to comport with the research pretty well and better than I was expecting.  It is surprisingly similar in it's general shape and slope on an excel graph, whatever that means. Maybe I just got lucky. It wouldn't be the first time.  

2. If I continue to use my own custom MC simulator should I now try to make it even more sophisticated on spending declines to reflect some of this research? I already have my own rudimentary assumptions baked in.

No, I don't think so.  In general, I am not very supportive of the over-use of Monte Carlo simulators in financial planning except as a very, very broad indicator of being on or off track or as a research tool.  In addition, I think that any "return" in decision making value from fine tuning my simulator would probably be zero in exchange for a lot of unnecessary work on a tool that is already close enough anyway and for a tool about which I have suspicions on its utility. 

3.  Is this declining spending approach even constructive for me at this point from a planning point of view?  I have been using something similar in a naïve way, but does it or should it mean anything?

Maybe.  At a minimum this idea appears to model observed behaviors in a large "pool" of retirees which I guess is something to think about.  Whether it makes sense on an individualized basis is something I don't know yet.  On the other hand it does, as I set out seven or more years ago to do, allow me to not over-estimate spending and that helps me tamp down on spend and fail rate fears within an uncertain future.  That, in turn, allows me more shoulder room on my "actuarial balance sheet" to explicitly budget for things like LTC and health shocks which are things a retiree maybe really should fear.  In the end, even though I seem to think this way on paper and in blog posts, I guess I'm not sure I totally believe the spending decline thing yet (except when looking at large pools of people) and anyway I already squeezed out a large 50% reduction in spending. I'm not sure how much more I can really squeeze out again at age 68 or 85.  I'm sure I will, though. 

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

[1] let's call it 2009 when I formalized it; it's been largely unexamined until today

[2] like I tend to do for example