A Decision Analysis approach to SWR choices

[Rich_in_Tampa]

I've spent a big chunk of my career in the decision analysis part of the health care world. The discussions here on SWR often remind me of how people can make tough medical decisions. I'd like to know how this admittedly technical approach works for you:

• There are two sides to each decision parameter, a "frequency" component (e.g. likelihood of maintaining your income til you die, a la Monte Carlo analysis), and an "impact" component - the subjective, personal, values-driven impact of having to live on a lot less income if things fail. You can assign each component a score between 0 and 1 (works better than % due to the arithmetic).
  
• The easy part: for a given SWR, estimate the likelihood of success using Firecalc or a similar calculator. This is basically a statistical probability* based on evidence, not subjective at all. So, if at 4% WR your probability of maintaining your adjusted income forever is 99%, the score is*0.99, 95% might be 0.95, etc.
• Now the hard part: assume your best case scenario for retirement quality of life gets a perfect score of "1." Estimate your quality of life compared to the best case scanario* if you had to take a serious cut in income half way through your anticipated retirement. A score of 1 means no affect on your quality of life at all: plenty of padding, just a number in your monthly statement, living well below your means to begin with. A low score (destitute, bad lifestyle, bills, whatever) might get a 0.2, and a tough but acceptable impact might get a 0.5. You be the judge, just use the same number when comparing different scenarios to one another.
• Assigning a number to this doesn't make it any less subjective in nature, but it forces you to come up with some type of reasonably consistent and useable number so you can compare your options.*
• Finally: multiply the two, e.g. 0.9 (firecalc success rate) x 0.5 (hypothetical quality of life given a serious income reduction if you fail) for a score of 0.45.
  
  Compare this to an alternate scenario of a 5% withdrawal rate, given a Firecalc success rate of .87 but where you are more comfortable handling an income reduction half-way out, say for an impact score of .75. The score here is 0.65 (maybe you chose to live well beneath your means, or could easily down size from a nice but way-too-expensive house, location, etc.). This strategy would be the winner for that example.

This approach might make it easier to understand the 4% rule in your own personal context. Others have made this point more clearly than I could, but here is a way to do it with numbers.* Depending on your circumstances, maybe a 5% rule is OK; for others, maybe 3.5%, and for still others, maybe you had better reduce your cost of living more than you initially thought, but go with 5% instead of 4%. There is no right or wrong (as far as the impact number), just whatever reflects your values.

Just a tool, but it can have a powerful effect on how you approach your decision for some people. A typical first reaction we hear is, "you can't put a number on something like...." but give it a try if you find it interesting, remembering that it is not the absolute number that counts, but rather its consistency across different scenarios you might be comparing.

[Martha]

Interesting and helpful.


My problem is in evaluating the risks and effects of failure. It would be one thing for us to have to reduce my spending by half because we lost half our assets and another thing to lose 90% of our assets. My problem may be overweighing the risk of losing more than half of our assets. So I end up seriously uncomfortable with a 4% withdrawl rate.

[Rich_in_Tampa]

Originally Posted by Martha

Interesting and helpful.

My problem is in evaluating the risks and effects of failure. It would be one thing for us to have to reduce my spending by half because we lost half our assets and another thing to lose 90% of our assets. My problem may be overweighing the risk of losing more than half of our assets. So I end up seriously uncomfortable with a 4% withdrawl rate.

Yes, it is hard to get a "handle" on risk, and we all tend to erroneously overestimate the likelihood of failure when the consequences of failure (impact) are high. It is best to keep them separate - you can think more clearly about it.

My take is that total poverty is not a plausible scenario (barring a doomsday scenario in which case we are all SOL). Rather, we will/should see it coming long before we run out of money, and adjust accordingly. That is, after 10 years we realize our asset base is down 10% from startup, so we will urgently reduce our expenses. This is why the Decision Analysis example is set up the way it is.

Anyone who sees they are down 10% real dollars after several years and doesn't make adjustments is destined for disaster. Haven't "met" many here who would allow that to happen.

[Martha]

Originally Posted by Rich_in_Tampa

Yes, it is hard to get a "handle" on risk, and we all tend to erroneously overestimate the likelihood of failure when the consequences of failure (impact) are high. It is best to keep them separate - you can think more clearly about it.

My take is that total poverty is not a plausible scenario (barring a doomsday scenario in which case we are all SOL). Rather, we will/should see it coming long before we run out of money, and adjust accordingly. That is, after 10 years we realize our asset base is down 10% from startup, so we will urgently reduce our expenses. This is why the Decision Analysis example is set up the way it is.

Anyone who sees they are down 10% real dollars after several years and doesn't make adjustments is destined for disaster. Haven't "met" many here who would allow that to happen.

I just had a "duh" moment. I can't keep working to simply reduce the risk of a doomsday senario. Nothing will reduce that risk.

[cute fuzzy bunny]

Many of the people worried about the "big problem" of a major financial meltdown, a nuke going off in new york, world war three, etc miss that fine point martha...

Like I always say, we dont have to beat/avoid the big disaster, we just have to 'beat' the other folks also running from it. No matter what happens, a well diversified investor with solid net worth that has made good decisions may lose 80-90% of what they 'had', but still be a good bit ahead of their contemporaries.

In short, you dont have to outrun the bear, just the other guy.

[Rich_in_Tampa]

Originally Posted by Martha

I just had a "duh" moment. I can't keep working to simply reduce the risk of a doomsday senario. Nothing will reduce that risk.

Ooohh - exciting . I have lots of "duh" moments - even weeks .

You are exactly right. Diminishing returns: biggest risk is missing out on all the fun that is 97% likely because you were working so hard to get to 99%.

[free4now]

Originally Posted by Rich_in_Tampa

Anyone who sees they are down 10% real dollars after several years and doesn't make adjustments is destined for disaster.

Well, anyone who retires into a bear market may be down much more than 10% after several years, but if they are withdrawing less than 3-4% they still have a ~100% historical chance of lasting a few more decades.

So it's a bit extreme to say people are destined for disaster if they don't make adjustments. The real issue is what adjustments can you make.. reducing your withdrawls from 4% down to 2% only puts you 2% ahead, which won't bring you anywhere near to your original real account value if you are down say 30% in real dollars in a few years.

Just a minor quibble... your reasoning overall seems quite sound. We all focus mostly on the worst cases here, but focussing on the best and medium cases is also very useful for planning purposes and your approach forces us to think about them as well.

Originally Posted by Cute n' Fuzzy Bunny

Many of the people worried about the "big problem" of a major financial meltdown, a nuke going off in new york, world war three, etc miss that fine point martha...

Like I always say, we dont have to beat/avoid the big disaster, we just have to 'beat' the other folks also running from it.
In short, you dont have to outrun the bear, just the other guy.

TH,

Can you outrun a nuclear weapon?

[HFWR]

Of course, we could be "nuked", but more likely by a dirty bomb than a warhead...

[cute fuzzy bunny]

Originally Posted by Cut-Throat

TH,

Can you outrun a nuclear weapon?

Yep, by living somewhere nobody would ever bother to nuke.

Although that might improve the appearance of some parts of this area.

Its funny, but the risk component of this vs the income makes sense to me. Thats why we have no debt and dont commit ourselves to any extensive out of pocket expenses. We can live comfortably off my wifes take home pay, and even she loses her job, we can slide by on dividends alone for a while without suffering too much. So my risk tolerance is pretty high.

[haha]

Compare this to an alternate scenario of a 5% withdrawal rate, given a Firecalc success rate of .87 but where you are more comfortable handling an income reduction half-way out, say for an impact score of .75. The score here is 0.65 (maybe you chose to live well beneath your means, or could easily down size from a nice but way-too-expensive house, location, etc.). This strategy would be the winner for that example.

This approach might make it easier to understand the 4% rule in your own personal context. Others have made this point more clearly than I could, but here is a way to do it with numbers

Rich, I understand your algorithm, but not your reasoning. If 2 people have the same starting $, use the same investments, experience the same economic environment, but one uses an SWR of 5% and the other uses 4%, the 4% guy is going to have more money anywhere along the trip. The 4% guy is
less likely to have to go to spending levels that he cannot tolerate than the 5% guy. In addition, his cutbacks in outlays should be subectively less onerous as well as objectively smaller. Many psychological studies have shown that people like to progress; they don't like to retrench.

It seems to me for example that the safest as well as probably the happiest approach for one with enough money to meet his resonable spending needs starting into retirement would be to spend very conservatively, thus allowing room for increases in spending if things go well, but probably not demading decreases unless things go very badly.*

There are some people on this board who advocate starting out spending more so you have a buffer between your usual spending and your rock bottom survival spending. To me this is hard to understand. For a given individual in a given environment rock bottom is rock bottom, and it is likely to feel worse coming from above than below. THe difference comes in becasue these people usually have a larger stash. So if* a5% SWR damages a $3million portfolio, he still has a pretty good portfolio, and he can live more like other people with less. This has nothing to do with his SWR except possibly negatively, rather it is a function of his wealth.

Where I can understand the use of your technique is in evaluating one's proper risk profile. If a wipeout on a promising investment would have a truly meaningful effect on his level of happiness, he shouldn't do it-even if success would allow him to spend a lot more time at Cabo. Putting numbers (utilities) to this would help to avoid obsessive and unhelpful ruminating.

I hope I have made my ideas clear; and I hope you will comment on them. BTW, I think you are a very helpful poster.

Originally Posted by Martha

My problem is in evaluating the risks and effects of failure.* *It would be one thing for us* to have to reduce my spending by half because we lost half our assets and another thing to lose 90% of our assets.* *My problem may be overweighing the risk of losing more than half of our assets.* So I end up seriously uncomfortable with a 4% withdrawl rate.*

Martha, how could you manage to lose half or more of your assets? Carry an umbrella and health insurance. Carry half of your stash in cash to medium term fixed. Then even if you lost 75% of your stock portfolio, you still would have better than 60% of your starting total portfolio left. And it is probably pretty hard to lose 75% of* a diversified stock portfolio.

Ha

[Rich_in_Tampa]

Originally Posted by HaHa

Rich, I understand your algorithm, but not your reasoning. If 2 people have the same starting $, use the same investments, experience the same economic environment, but one uses an SWR of 5% and the other uses 4%, the 4% guy is going to have more money anywhere along the trip. The 4% guy is less likely to have to go to spending levels that he cannot tolerate than the 5% guy. In addition, his cutbacks in outlays should be subectively less onerous as well as objectively smaller. Many psychological studies have shown that people like to progress; they don't like to retrench.

It seems to me for example that the safest as well as probably the happiest approach for one with enough money to meet his resonable spending needs starting into retirement would be to spend very conservatively, thus allowing room for increases in spending if things go well, but probably not demading decreases unless things go very badly.

Ha,

If you would be just as happy with $400 per month as with $500, of course $400 is the best strategy since it is more of sure thing long term. However, there is some number at which you would be less happy than at $500; maybe it's $350 or $300. Whatever that number is, we all have it. Now you need to compare the happiness at $500 x the likelihood of lifelong asset preservation using $500, with the happiness at $350 (or whatever your "now I'm not so happy" number) times its higher likelihood of long term success:

a. $500 per month: .8 likelihood of success x 1 (ideal happiness scenario) = .8
b. $350 per month: .98 likelihood of success x .7 (not as happy) = .67

Choice "a." wins here. Now let's run it with your implied example, where you are almost as happy at $350 compared to $500:

a. $500 per month: .8 likelihood of success x 1 (ideal happiness scenario) = .8
b. $350 per month: .98 likelihood of success x .9 (almost as happy) = .88

Choice "b." wins here.

I hope this helps - you can decide how much unhappier you will be at the lower income, and whether the trade-off in certainty of asset preservation is worth it. Again, just a tool and all decisions it generates are right or wrong only for you, not in any absolute sense.

[brewer12345]

Rich, this is more of a general question on your algorithm, but how does one decide what is a high enough number to make an affirmative choice? Clearly in a medical situation where you *have* to do something because otherwise somebody will drop dead, using decision analysis makes sense because you can rank order your optons and then make the best choice. But if we are talking about the decision to retire, one (usually) does not have the same "burning platform" since the would-be FIREee can just stay at work. So I am having a tough time making the leap here...

[Rich_in_Tampa]

Originally Posted by brewer12345

Rich, this is more of a general question on your algorithm, but how does one decide what is a high enough number to make an affirmative choice? Clearly in a medical situation where you *have* to do something because otherwise somebody will drop dead, using decision analysis makes sense because you can rank order your optons and then make the best choice. But if we are talking about the decision to retire, one (usually) does not have the same "burning platform" since the would-be FIREee can just stay at work. So I am having a tough time making the leap here...

Great question. In the "real world" you would compare harm to benefit. If Harm:Benefit is greater than or equal to 1, you pass (i.e. decline treatment); otherwise you accept treatment.

Here I suggested it in in the limited context of simply comparing two sets of post-FIRE circumstances and deciding which was better. To answer your question at the risk of boring you to death:

1. Perform the exercise using your current pre-retirement lifestyle. Use a likelihood of 1, since this is the actual current state of affairs. For the "subjective" part, look at your happiness level, security, social networking, income, work hours, etc. and come up with a quality of life score from 0 to 1, just as discussed.

If your pre-retirement life is peachy, secure, and you feel great about the state of your world, have enough free time, etc.* it might approach a score of 1. We all know people who seem to be there. However, given that this is an early-retirement forum, I suspect this will not often be the case.

2. Consider each of your post-FIRE options and perform the exercise with each one.

3. Look at the scores, including your pre-FIRE score. If pre-FIRE beats all the* post-FIRE scores, keep working. You are lucky. If any of the post-FIRE scenario scores is higher than the pre-FIRE score, it's time.

Of course you have to be realistic in your assumptions and judgments, and avoid the human tendency to "game it" so that you retire before you are ready. But that's a whole other discussion.

Hope that helps.

[MasterBlaster]

I posted something along these lines yesterday suggesting that the OPs model was not strictly valid but deleted it so as to not be too critical.

Anyway the OP suggests that if you can put quantitative numbers on the SWR_success and the happiness_factor that the joint probability for these two factors that the OP calls the quality_of_life is just the product of the two. If the happiness_factor and the SWR_success are indeed statistically independent then the multiplication model will exactly discribe the joint probability for "quality of life".

In my opinion the happiness factor is highly dependent on the SWR_success and that the two are highly correlated. This therefore violates the statistical independence criteria required to determine the "quality of life" joint probability using a multiplicative model. Furthermore while the SWR_success is (arguably) determinable provided past pertformance is indicative of future performance, the happiness_factor being derived from feelings is not so determinable.*

Therefore, The OPs model may give some insight into how someone feels about their retirement but is along way from an analytic model of retirement success.

I hope I haven't been too critical here. These type heuristic models are challenging.

[Nords]

Originally Posted by Martha

My problem is in evaluating the risks and effects of failure. It would be one thing for us to have to reduce my spending by half because we lost half our assets and another thing to lose 90% of our assets. My problem may be overweighing the risk of losing more than half of our assets. So I end up seriously uncomfortable with a 4% withdrawl rate.

Perhaps that's a diversification issue. We may have more than 90% of our retirement portfolio in equities, but we also have a govt pension and a rental property. You seem to have quite a few defensive layers in your retirement portfolio, too, notwithstanding Greg's attitude.

Originally Posted by Cut-Throat

Can you outrun a nuclear weapon?

I can assure you that the Russians & Chinese, even Al-Quaida, would have a hard time justifying the waste of a nuclear weapon on Travis AFB, let alone Yuba City. How would you tell whether there was battle damage?

I'm not so sure what Minot & Omaha have pointed in that direction, however...