Tuesday, March 3, 2015

The Role of Science in Retirement Planning

IMO, judgment, common sense, creativity, etc. are the most important skills of the financial advisor; the more common pursuit of "science" may be a fool's errand.  One group advocating science suggests MCS modeling that ignores black swans.  Another group of scientists suggests much of this is pseudo science, and, among other things, demands inclusion of black swans (usually leading to safety-first for all).  A third group attempts to straddle the above two positions. 

I am not a CFA and I do not have a PhD in math or economics.  What I do have is a JD, CPA, and many years of analyzing and resolving complex international corporate tax and financial structuring projects.  There were always plenty of "scientists" (experts) willing to sell me the latest and greatest product for dealing with these projects (at $1m+ price tag).  Instead, I always started each project with a blank slate and built up the analysis from scratch; making sure I did not overlook any relevant and material issues.  In retrospect, there was never an appropriate cookie-cutter solution, even when offered by the world's greatest experts at their high price tag.  The same blank slate approach should be applied to retirement planning.

I am not suggesting that knowledge is unnecessary.  Knowledge is essential.  The BFP article does an excellent job of identifying expected future returns, and that knowledge is crucial to retirement planning.  But in addition to knowledge, the most important  skills are judgment, common sense and creativity.  The advisor should start with a blank slate and make certain he understands all of the relevant and material information of the client.  Perhaps most importantly, the information gathering must make sure that the advisor and the client are on the same wave length regarding spending.  I found that the only way to make sure there is a complete understanding on spending is to test relevant AAS scenarios with the client, as discussed in my earlier post.  It is up to the judgment of the advisor to decide which scenarios are relevant and important -- no science can do that for you.  There is no science that can give you exact probabilities of the chosen scenarios, again it is a matter of judgment.  IMO, good judgment will result in estimated probabilities that are as good as probabilities that are derived from 10,000 iterations  (which, perhaps, conveniently ignore black swans or autocorrelations).  Of course, all analyses must be regularly updated to take into account changes in facts and circumstances.

Saturday, February 7, 2015

A Little Thought Experiment



Take a look at two of the articles and discussion topics on APV -- "Retiring in a Low-Return Environment" and "How to Link Retirement Strategies to Sustainable-Spending Rates".  Both are a testimony to the never-ending quest for the best scientific approach to optimal retirement planning.  Both articles are intelligent and logical and dutifully report the results of volumes of research and analysis.  I commend Messrs Blanchett, Finke, and Pfau (BFP) on their diligence and fine work.  Of course they are only three of many who have contributed to the vast volume of research on retirement planning.

OK, now that the niceties are out of the way, let me turn to my thought experiment.  Let's begin by scrapping all existing models and formulas that are used for retirement planning.  No MCS,  SWR, glidepaths, guardrails, etc.  Instead we are going to ask BFP and perhaps others, say, Bill Bernstein, Swedroe, (and maybe even Edesess and Taleb, if they promise to behave themselves :-) ) to put their heads together and see if they can agree on two things -- expected returns and reasonable worst case scenarios.

By expected returns I simply mean the stock and bond expected returns going forward, ignoring sequence of returns, outliers, etc.  BFP et al (team) will no doubt consider current stock market valuations, current interest rates, global market conditions, etc. and come up with what they believe are fundamental expected returns going forward.  I suspect that it would be relatively easy to come up with agreed expected returns within a narrow range.  I further expect that the team would also find it relatively easy to agree such base case expected returns at any time in the future, adjusting their estimates to the then existing market conditions.

Agreeing reasonable worst case scenarios will be a tougher challenge for the team, but I suspect that again agreement can be reached within a fairly narrow range.  Both market and inflation scenarios should be considered.  Presumably reasonable worst case market scenarios will go beyond the general periodic sequence of returns fluctuations.  Some of these can be quite severe, e.g.2008, but relatively rapid reversals take them out of worst case.  The question would be: what is a worst case scenario that you believe would be reasonable for a retiree to prepare for.  Peoples' imaginations could go to imploding into black holes, but when considering reasonable worst cases, I suspect that something much less drastic would arise.  I suggest starting the search with an immediate permanent equity decline of 60%; not too unlike Japan.  Then each member of the team could submit his views on whether a reasonable worst case should be more or less severe.  I suspect that a narrow range of reasonable worst case scenarios could be reached.  A similar exercise would be necessary to arrive at a reasonable worst case inflation scenario and any other type of worst case that might be considered .

OK, now we have agreed expected returns and reasonable worst case scenarios, now what?  Unlike MCS, we have no idea of the probabilities of the worst case scenario occurring.  Obviously many other events between base case expected returns and worst case might arise, and we have not identified their nature or likelihood of occurrence.  What good are two points when there is an infinite number of points in between?  Time for the thought experiment.

My hypothesis is that the two points -- base case and worst case -- are the only forecasting tools that are needed for retirement planning.  The endless MCS studies, etc. can be scrapped and replaced with these two items.  If this hypothesis is correct, then it is obvious that retirement planning technicalities will be substantially reduced and simplified.  The base case is our 50% point, and we know our actual could be more or less, likely centered on that point.  We know that our worst case is possible, but very unlikely to happen.  We know that other less worse events between base and worst may happen, which may become more likely as we move from worst to base, e.g., a 30% permanent equity decline is more likely than 60% and a dip of short duration is more likely than a permanent decline.  Starting with the worst case, it is easy to evaluate possible less worse cases and obtain a complete picture of what the future possibilities may be.

"But we have lost MCS that derived all those in between probabilities."  Really?  Those probabilities were simply unreliable and therefore unhelpful.  But MCS told me that I had an 85% probability of success in meeting my goals.  Yeah, but what if the worst case of a 60% permanent (or even 20 year) equity decline happened tomorrow?   "But my MCS and guardrails told me the amount I can spend."  Unfortunately, that will not work out for you given the long-term decline.


I am not suggesting that retirement planning should be geared to cover the worst case, which is extremely unlikely to happen.  What I am suggesting is that once we have both base and worst, then it is possible to evaluate spending and investing by considering any of the outcomes that fall between those two, which the advisor considers most pertinent and relevant to the client.  For example, the advisor could show the client how an event like 2008, which is definitely possible, would impact spending and investments.  The extensive computer modeling of MCS, etc. is replaced with planning for various contingencies that are most relevant to the client's situation.  Most importantly, the client understands what a worst case is, and how other less worse case scenarios would play out in his specific situation.  Judgment replaces formulas.  Needless to say, the base and worst case are continually reviewed and revised as necessary, together with the rest of the retirement planning.

Tuesday, July 29, 2014

Spending Levels in Retirement

The article focuses on the range of annual available spends during retirement.  Here are some thoughts on how the spending levels might affect investments.

How much will you need in retirement, and how does investing tie into this?  To what extent might (should) a retiree be willing to risk not reaching planned spending levels?  Consider these three levels of spending.

               Spending Level 1.  A sustenance level where the individual or couple lives healthily with little more, say 40-50k.

               Spending Level 2.  Minimum happiness level.  Daniel Kahneman and others have concluded that there is a spending level above which additional income does not bring more happiness, say 80k.

               Spending Level 3.  A level above Level 2 that maintains (or exceeds)  the pre-retirement lifestyle, say 100k or more, with no upper limit.

Most retirees will fall somewhere in the first two levels.  How much safety is required at each level?  Considerable conservatism, perhaps using a liability matching portfolio of annuities or a TIPS ladder, will likely be optimal at Level 1, and some similar degree of conservatism may be desirable at Level 2.   No rules seem appropriate for Level 3.

Richard Thaler and many others have written extensively on the behavioral aspects of investing.  The behavioral aspects of spending don't seem to have received much attention.  Some at Level 3 are extremely frugal and others are spendthrifts.  Some need to keep up with the Joneses, others could care less.  Some are flexible in lifestyle and others inflexible.  Some have an irrational fear of running out of money, and some are eternal optimists.  Even at the level of super wealthy, we have Warren Buffett and Dennis Kozlowski.


For those at Level 3 it is likely unnecessary and too costly to create a LMP to assure certain spending availability.  When considering all of the behavioral differences, it seems clear that there is no single formula that will apply to all.  At Level 3, all planning must be ad hoc.  

Thursday, May 15, 2014

The two main criticisms of Annual Available Spend

The two main criticisms of this methodology are: 1) this is nothing new.  Most everyone does the initial information gathering and analysis using Excel models or the like; and 2) the methodology does not come up with a safe withdrawal rate or similar plan for withdrawals.

            Routine information gathering, or something more?

The main premise of the AAS methodology is that information gathering is the most important aspect of retirement planning.  Everyone does something in this regard, but often not enough for effective planning.  If there is a complete understanding of the retiree's specific facts, the planning will be centered on those facts and not simply follow one of the many formula's currently in use.  So do most gather information in a manner comparable to AAS?  The answer is yes if the answers to the following three questions are yes:
           
            1.  Do you know your base case AAS:  The after-tax amount received from all sources that will be available for you to spend each year until your assumed year of death (age 100 or ?) when investments reach zero?

            2.  Do you know your AAS computed under a reasonable worst case:  For example, equity value permanently declining by 60% (or more or less severe market decline, or high inflation); or whatever you consider to be a reasonable worst case as applicable to your specific situation?

            3.  Have you compared your base and worst case AAS to your best estimate of expected future annual spending needs?

As the article demonstrates the computations of best and worst case AAS and spending needs are prepared in great detail to make sure all of the retiree's specific information is properly taken into account.  It is my understanding that information gathering generally does not include computing base and worst case AAS, and spending needs may or may not be thoroughly evaluated, e.g., they may not obtain the retiree's detailed view on how he might practically and psychologically deal with various negative scenarios.

In cases such as retirement planning, when analysis and problem solving involve matters of uncertainty, information gathering is more than half the battle.  So after we have this information, what do we do?  That brings us to the second criticism.

            So where is the method for determining investments or SWR?

Once the information gathering is completed, there is no specific investment or withdrawal plan that fits all.  The facts will lead to the best approach using solid reasoning and judgment. The base and worst case AAS are compared to spending needs and that is where the important analysis lies.  Is the retiree more like Joe, Bill, or Al in the article?  Of course, no two individuals are alike and it is impossible to generalize.  Are the AAS scenarios more or less than spending needs?  To what extent can the retiree manage shortfalls or a worst case?  What is the psychological position of the retiree?  Etc. 

Changes in investments can be tested in the AAS best and worst case models -- fixed/equity allocations, annuities, TIPS ladders, etc.  Each of these will change the base case and worst case AAS, which can then be compared to actual spending needs.  Various withdrawal plans or SWR plans can also be tested and compared.  There are many systems involving various SWR rates, value timing, buckets, reservoirs, etc.  If any of these might be helpful, test them in the models, but there is no best system or formula; it is all dependent upon evaluating the AAS model outcomes.


One additional point that must be reemphasized.  It is essential to continually update and monitor the model output.  The model can easily and automatically update the base case AAS.  In the short term there should be very little change in the AAS, and this might lend comfort to the retiree.  However, some changes are inevitable over the years (hopefully for the good), and adjustments can be made as appropriate.

Friday, May 2, 2014

Summary and Rationale for Retirement Planning Using Annual Available Spend

There have been many attempts to create a science of retirement planning.  The premise of this article is that no effective formulaic approach exists.  Instead of relying on such techniques, it is incumbent upon us to thoroughly understand our present financial position and then evaluate how we will manage our finances given a range of possible future outcomes

It is, of course, possible to determine odds of success if we assume that one's investments will exactly follow the past.  Although we can learn from the past, "scientific" reliance on the past has been demonstrated to be wrong.  For example, few if any people actually believe the 10% equity returns of the past will be repeated.  The types of market declines and inflation of the past are a rough guide for the future, but we should not place meaningful reliance on them.

Perhaps the most common ''science" is safe withdrawal rates, SWR.  In the past it was widely accepted that the Trinity Study proved that a 4% withdrawal rate was safe.  Today with the advancement of the SWR science, we deal with a range of <2% to >5%.  With this broad range, and the many alternative methods of implementation, SWR is of questionable value.  Having said that, SWR can be a useful tool for those who want ballpark future planning, e.g., if I save 1m, I will have about 30 or 40k of cash flow when I retire.

The objective of SWR is to determine the amount of annual spending that is "safe".  If safety is the goal, there are methodologies that can be used.  Life-cycle investing and Liability Matching Portfolios (using, for example, TIPS ladders) are used to eliminate or minimize market and inflation risk.  These approaches are the best means of achieving safety, but at a very substantial cost.  Some individuals do not have a large enough portfolio to live off the low returns generated by this approach.  Those who do must sacrifice a great deal of upside potential.  Not using such safe approach does not mean the alternative is unsafe.  Instead it will mean accepting a small possibility of some (acceptable) reduction in retirement spending in exchange for a much larger likelihood of increases in available retirement spending.

Monte Carlo Simulation is a popular scientific approach that is used to evaluate the variability of future returns.  MCS is used to estimate probabilities of success of reaching one's retirement goals based upon 1000's of computed iterations.  The outputs from MCS are, of course, only as good as the assumptions that are input.  Faulty inputs, often using historical facts, have resulted in corresponding false probabilities of success.  Although good inputs can result in a meaningful evaluation of odds of success with respect to sequencing of returns, MCS fails to deal with Nassim Taleb's Black Swans or deal with Bill Bernstein's caveat " any estimate of long-term financial success greater than about 80% is meaningless".  Inclusion of outliers in MCS defeats the purpose of the science.

Some financial planners prescribe safe approaches such as life-cycle or liability matching portfolio methodologies, but at the cost of very low returns. The majority of financial planners use one or more formulaic approaches discussed above, which are regularly revised and expanded with the ever changing "science".  The quest for certainty by advisors and their clients is understandable, but no formula can provide that certainty. 

Because there is no effective scientific approach to retirement financial planning, it is incumbent upon us to thoroughly understand our present financial position and then evaluate how we will manage our finances given a range of possible future outcomes.  The methodology described in this paper is limited to establishing these fundamental facts and analyses.
  
The first year law student is told that the most important factor in earning credit on exam questions is to demonstrate an understanding of the relevant and material facts.  That is the best advice for anyone who is attempting to analyze and solve problems where there is no certainty.  When I began my own retirement planning, I researched the subject thoroughly.  I could do a reasonable job of estimating my range of spending needs in retirement, but I realized that nowhere did the retirement literature deal with my most basic question:  How much will I have available to spend each year for the rest of my life?  Of course, given the uncertainty, there is no precise answer to this question, but it is possible to determine a reasonable range for retirement financial planning.


This article explains the methodology for assembling the relevant and material facts necessary for effective retirement financial planning.  It does not pretend to assess the probabilities of various worst case scenarios.  It does not attempt to evaluate the risk inherent in various investment portfolios.  Instead, it provides those facts, specific to the individual, that are essential to set the stage for effective retirement planning by the retiree or his advisor.

Wednesday, March 26, 2014

Retirement Planning with Annual Available Spend

The retirement planning methodology and model described in this article were developed over a few years to assist me with my retirement planning.  I am a retired tax attorney and CPA.  My career was in international tax planning; first as a tax partner with a major international CPA firm, and next as a VP Tax for a large multinational corporation.  I am now retired with no aspirations to become a financial planner.  I have been using this model for four years, and it provides what I need to plan my retirement; I hope that others will also find it helpful.
My general investment philosophy is passive, and it generally follows that of Jack Bogle, Bill Bernstein, and Larry Swedroe.  These individuals generally agree with Bernstein's caveat in the Retirement Calculator from Hell, Part III that: any estimate of long-term financial success greater than about 80% is meaningless.  Given this uncertainty, my approach relies less on the "scientific", and more on judgment applied to a thorough understanding of the client's facts.
My methodology does not incorporate the commonly used SWR or MCS methodologies, nor does it rely on the past in order to predict the future.  Given the uncertainties of the future, I focused on how much I would be able to spend each year for the rest of my life, given fundamental expected returns and inflation, and then considered how negative and "worst" case scenarios would affect my retirement planning.  That is the basis for the Annual Available Spend (AAS) methodology described in this paper.
My initial focus was solely on my specific situation, so I was particularly careful to obtain all relevant and material information.  I believe that detailed focus on the specific client situation is also essential.  From this point forward I will refer to advisor and client.  
General Framework of Methodology
1.  The advisor will work directly with the client to determine all parameters of his expected spending needs during retirement.
2. The advisor will obtain a complete analysis of all sources and types of the client's income.  Each of these items will constitute a different column in the model.  All items will be entered after tax, so it is necessary to have a full understanding of the client's income tax position.
3.  All information will be entered into the model, and fundamental returns and inflation applied.  A conservative year of death will be included.  The model will then compute base case annual available spend -- the amount that can be spent each year, with investments decreasing to zero in the year of death.
4.  Negative and "worst" case scenarios will be entered into the model.  The scenarios chosen will be keyed to the specific situation of the client.
5.  The AAS Excel model can be set to automatically adjust for market value changes and to automatically recompute base case AAS.  Cash and portfolio quantities can be updated monthly.  The advisor must determine if and when the base case and/or negative assumptions should be revised.
6.  The advisor will review the AAS scenarios with the client, and evaluate the client's spending and investment plan on a periodic basis.  Changes will be made as necessary and appropriate.
1. Spending Needs
Determining the retirement spending needs of the client is, of course, a routine exercise in retirement planning.  The client should keep a monthly record of expenditures, with or without a detailed budget.  Large or exceptional items should be noted.  As part of this methodology, income taxes should be eliminated from annual spending totals since they are netted against income in the model.  Of course the more prior years included, the better the feel for required and discretionary spending.  Consideration should also be given to unusual and contingent future expenditures.
Just as clients differ with regard to investment risk tolerance, they will differ with regard to spending tolerance.  Some may feel very strongly about maintaining certain minimum living standards, while others may be much more flexible regarding their ability to cut back if the need arises.  Attention to these and other behavioral issues is essential.
2. Income Analysis
In order to properly evaluate the client's situation, it is, of course, necessary to obtain a full analysis of all types of client income  -- pensions, annuities, deferred comp, Soc Sec, taxable investments, tax deferred investments, Roth IRAs, RMD, etc.  Each of these items will be included in a separate column of the Excel model.  Income taxes are netted against the various income items.  The estimated taxes need not be precise, but must be materially correct.  It is necessary to compute both base tax and marginal tax.  The acid test will be to make sure that the total differential between all items of gross income and the after-tax income included in the model is materially the same as taxes computed using TurboTax or similar application.
It will be necessary to program the model to properly source the income for spending from the appropriate pots of income.  Typically this will first come from def comp, pensions, annuities, and social security, followed by RMD, taxable investments or Roth, tax deferred, etc.
3. Base Case AAS Model
Following is an example simplified base case AAS model:
This model establishes a base case for Bill.  In this example all investments are included in a single column, and the model assumes Bill has a 50-50 equity fixed split.  This simplified example does not include a column for tax deferred investments or RMD.  Each type of after-tax income is included in a separate column.  The three assumptions for the base case model are after-tax return on investment, inflation rate, and year of death.
The model does not assume historical investment returns.  Instead it assumes the widely used "fundamental" returns that consider current market valuations.  There seems to be a broad general consensus regarding what these returns should be.  For purposes of illustration, it is assumed that the equity-fixed real returns approximate 5% and 1% respectively.  This would result in combined nominal returns of 5.5%, and roughly a 4% after tax return.  The inflation rate assumed of 2.5% also represents an approximate generally accepted current inflation rate.  The model increases annual expenditures each year by the inflation rate, and it increases the after-tax social security income by the inflation rate less .5%.  If Bill is now 70, his assumed year of death will be age 100.  Of course, advisors will differ on which rates and ages to assume.  A slight edge to conservatism in all assumptions will likely be desirable.  Tax effecting will vary depending on individual circumstances.
We then run an Excel macro that applies What if Analysis, Goal Seek to determine the amount Bill can spend each year with the investments declining to zero in the assumed year of death.  In this example Bill's Annual Available Spend is about 88k.  We can now compare the base case AAS to the 85k of spending needs; so-far-so-good.
Note this and the next examples deal with a single individual and ignore the spouse.  A later example will consider the surviving spouse situation.
4. Negative Scenarios -- The Planning Begins
Up to this point the advisor has been assembling all the necessary information to begin the retirement planning.  Now the planning begins.  All will agree that computation of the base case AAS cannot be the end product in itself.  Now we must decide how to deal with sequencing of returns, black swans and the like.
There are various approaches to retirement planning.  Many prefer approaches that are more actuarial or "scientific".  All such approaches have validity, but considering the inherent uncertainty that cannot be overcome, this methodology relies on the knowledge and judgment of the advisor coupled with a detailed understanding of the individual client's situation.
Once we have our base case AAS, we can factor in negative scenarios, which may seem at first blush to be an impossible task.  We might say we should prepare for the worst case scenario; however, "worst" is an absolute and there is always something worse until we get to the point where the earth implodes and disappears into a black hole.  Instead, we must attempt to come up with various negative scenarios including reasonable "worst" case scenarios.  Of course we will look to the past, not to determine what event might happen, but to get a feel for the degree and scope of the potential problems.  
Any choice of negative or worst case scenarios must be keyed to the client's specific situation.  Worst case market scenarios will not be particularly helpful for Bill, whose income is primarily from fixed pensions.  In his case increased inflation will be a much greater concern than market declines.  William Bernstein's discusses various types of risk in his recent book, Deep risk: How History informs Portfolio Design.  Using Bernstein's terms, there are both shallow risks, where market declines reverse in a relatively short time, and deep risk where declines may never reverse, or where there is high inflation or deflation.  
Although seemingly difficult, it may be possible to obtain a reasonable worst case general consensus among financial advisors.  For example, with respect to a reasonable worst case market decline, we might start by suggesting a permanent 60% decline in equities.  The 60% equity decline may not seem too severe, but coupled with permanency (as occurred in Japan in 1990) this. or something of this order, might qualify as a reasonable worst case scenario.  Would a group of advisors reach general agreement that this example is too severe, or not severe enough? Could a broad general consensus be reached?  I believe that it could.
Regardless of any consensus, the important task for the advisor is to pick various negative scenarios that are relevant and appropriate to the specific situation of the client.  Following are three simple case studies to illustrate this approach.  In each of these the market decline scenario assumes a 50-50 equity-fixed split, and that equities suffer an immediate permanent market decline of 60%.  This results in an overall investment decline of 30%.  In each case it is assumed that the client rebalances to 50-50 immediately following the decline.
Bill Example -- High Percent of Fixed Income
As can be seen from the Base Case model run above, the majority of Bill's after-tax income comes from fixed pensions.  Here again is the base case summary:
Two reasonable worst case scenarios were selected for Bill.  One is the 30% permanent decline in investments discussed above.  The second is an inflation increase from 2.5% to 5%.  Each scenario was independently entered into the model and a revised AAS was computed.  The inflation scenario inputs and AAS are shown below.  The results of the two scenarios are shown in the box on the right:

In step one, we assume that the advisor and client agreed an actual spend of 85k.  We can see from the market decline scenario result, that even in the "worst" case, the impact on spending is not substantial, from 88k to 83k.  However the effect is much more severe in the inflation scenario because of the fixed pension income.
Note that in the inflation scenario, the model adjusts nominal rates of both investment returns and Social Security increases in order to maintain real rates at the same level.  This scenario shows a 10k decline in AAS, with spending 8k below the initial actual spend.  Of course any number of additional negative scenarios are possible.  There is no formula available to deal with election of scenarios and analysis of their implications.  This is up to the skill and judgment of the advisor, considering the specific situation of his client.
Joe -- Moderate Investment Portfolio
Joe has a moderate portfolio of 600k, after-tax social security income of 30k, and no fixed pensions or other income.  Following are the base case summary and the worst case 30% permanent reduction in total investments summary:
The investment decline scenario reduces Joe's AAS about 8k below base case and 4k below actual spend.  Maybe Joe has little concern regarding spending and believes he could easily live on 46k.  Maybe he has considerable concern and feels the 50k is tight.  In the latter situation the advisor might explain the remote possibility of the worst case occurring and present alternative negative scenarios to Joe, as discussed below.  Maybe the advisor will consider spending adjustments or investment changes with his client.  There are many ways this might play out between advisor and client; in such cases the judgment of the advisor may be much more important than the "science" of alternative approaches.
The investment decline scenario assumes a permanent 30% decline in total investments (60% in equities).  This is intended to represent a reasonable worst case since the decline is permanent.  In the more typical case where the losses reverse in subsequent years, the reduction in AAS would be much less severe.  In Joe's case, if we assume that the after-tax investment return increases from 4% to 10% in years 2 to 5 following the decline (similar to the US 2008 decline and recovery) , Joe's AAS increases from 46k to 50k.
Al -- More than enough
Al has a portfolio of 3m, and he is not excessive in his spending habits, with actual spending at 110k.  He has only a moderate fixed pension, and his retirement income will be sourced primarily from his social security and investments.  Following is Al's base case AAS model run:
The first item of note is that the base case AAS of 163k substantially exceeds the actual spend of 110k.  This demonstrates that Al could, for example, gift about 53k (163-110) to heirs each year and not worsen his base case financial position.  The next item of note is the size of the estate available for heirs if Al does not live the full 30 years until age 100; at age 90 under the base case, Al's estate will approximate 2m.  Now let's consider Al's worst case scenario of a permanent 30% decline in investment assets:

This methodology can be particularly helpful in situations where the client has more than enough.  Often behavioral factors arise, and the client must be convinced that he can easily spend more.  For example, Al can make substantial gifts to children and/or grandchildren without fear of a serious shortfall in the future.  Even under the worst case scenario of a permanent 60% decline in equities, Al's AAS will still be almost 15k more than the actual spend.  Assurances of 100% odds of success in meeting one's goals may well be less comforting and convincing to the client than seeing the computations of how worst case scenarios will still leave him with an AAS greater than his actual spend.  The model is also helpful in demonstrating the minimal effect on AAS of increased spending, for example, purchasing an expensive new car.
Bill's Spouse -- Pension Reductions
The above examples do not consider the case of the spouse of the client.  In those cases where the primary sources of income are social security and investments, there should be little impact on the spouse if the husband predeceases.  In Bill's case most of his income comes from fixed pensions.  If we assume those pensions provide for a 50% survivor benefit on Bill's (H) death, the AAS will decline substantial for Bill's surviving spouse (W).  Assuming Bill dies at age 85 and his spouse survives until age 100, here is the AAS of the spouse:
In Bill's base case AAS model run we can see that in year 15 his actual spend had increased to 125k and his total investments had increased to 495k.  In actuality, after 15 years Bill's situation is likely to change materially and might not resemble anything close to the above.  However for initial planning purposes it is necessary to follow through the original plan in order to evaluate the relative effect on W in the event H predeceases.  In the above example, the spouse inherits the investments in year 16, but her social security is reduced to survivor benefits, and pension income is cut in half to 30k.  At Bill's death their joint annual spend had increased to 125k, but spouse's new base case is 94k, or approximately 76% of the couple's actual spend.  The high inflation scenario will not substantially change this, but the 30% investment decline scenario will result in an AAS decline to approximately 66% of the prior actual spend.  Declines of AAS to 76% or even 66% might seem be acceptable to H and W, or they might consider this problematic.  It is essential that the advisor explain the relative differences in AAS of the surviving spouse to the client and proceed as appropriate with the planning.

5. Continual Updating and Periodic Review
The Excel model can be set to automatically update values of investments and recompute the base case AAS.  Changes in quantity of investments and cash balances can be input monthly, together with updates to actual spend.  The client with access to such model will readily see that most market moves have relatively little impact on the AAS; similarly what may seem a large additional spend may barely budge the AAS.  The base case inputs are generally conservative, so it may well be that the client observes a continual gradual increase in AAS, which may allay some market fears.  
Once set up, the advisor can monitor the changes, but, typically will not need to adjust the model.  Of course, very major market shifts, up or down, might occasion a revisit.  The base case assumptions of expected returns, inflation, and life expectancy will not require frequent revisions.  Given the conservative nature of these assumptions, hopefully changes will more often be for better than for worse.

6. The Overall Plan -- Summary
As can be seen from the above, this methodology does not attempt to develop a formulaic or scientific approach to retirement planning.  The approach instead looks to obtain the best and most complete client information, and then compute a base case AAS using generally agreed conservative fundamental returns, inflation, and life expectancy.  The approach accepts the uncertainty of future returns.  It does not attempt to apply historic patterns or statistics, nor does it attempt to assess probabilities of success through modeling sequencing of returns.
The advisor will need to select negative and worst case scenarios that are most material and relevant to the client's situation.  Presentation of a reasonable worst case scenario is mainly to prepare the client for how bad it could be, in the context of AAS.  The advisor will explain that there is a very low probability of the worst case occurring; he may also wish to run less negative scenarios, like loss followed by reversal in the above Joe example.  
The client should easily understand the model outputs, since the AAS scenarios can be compared directly with his actual spend.  Joe is concerned with having enough and setting the right spending levels; Bill is concerned with inflation and reduced pension amounts available to his wife, and Al may need to understand that he has the ability to spend more than at present.
Advisor investment advice is beyond the scope of this article.  Advisor's recommended investments might range from passive to active, conservative to aggressive.  The AAS modeling can play an important role in making these investment decisions, and, importantly, explaining their impact to the client.
I have found this method to be very useful in my personal retirement planning, and I hope others will as well.

John D. Craig, J.D., CPA