Historical vs Monte Carlo
@nc-cpl for a pretty good description of setting up and managing a bucket approach see Fritz Gilbert. https://www.theretirementmanifesto.com/how-to-build-a-retirement-paycheck/ and https://www.theretirementmanifesto.com/how-to-manage-the-bucket-strategy/ . I was actually just reading them 🙂
@nevinsalumni-duke-edu Thanks Bob - The tricky part as I see it is this...PRC calculates an "order" of which accounts to tap in a specific order. In our case it shows Regular as first every time I run it, followed by TD's, then Roth. Fine. But if the bulk of our stock funds reside in our Regular" brokerage accounts, and those should be whats in Bucket 3, that runs counter to PRC's recommended drawdown order, right?
@nc-cpl PRC's recommendations are much more granular and strategic than the simple bucket methodology. But (big but!) PRC can't tell what will be happening in the future. So I'd always stop first and think about that before liquidating either equities or bonds, and what the immediate consequences would be in the current environment. For example, if the market was down, I wouldn't liquidate stocks at a loss, I'd look for another solution (cut expenses, use cash, etc). Unless of course it was in a brokerage and they were shares I wanted to get rid of anyway, and could take advantage of some nifty loss harvesting. You could always model this in PRC before doing it in the area where you specify manual disbursements.
Interesting new article by Kitces about Monte Carlo and Historical simulations:
Contrary to popular belief, Monte Carlo simulation can actually be less conservative than historical simulation at levels commonly used in practice. And while current financial planning software generally provides an adequate number of Monte Carlo scenarios, the deviation from historical returns at particular spending risk levels provides some additional insight into why multiple perspectives may be useful for informing retirement income decisions. Which suggests that incorporating tools that use a range of simulation types and data could provide more realistic spending recommendations for clients!
Right up PRC's alley 😊
@giovanelli766 I don't recall, but doesn't PRC run 500 simulations? Would this be making the case for 1,000 (but no more)? I realize it would add a bit more processing time, but if you're not running the analysis over and over again on a daily basis, wouldn't the time tradeoff be potentially worth it? Alternatively, maybe PRC could be configured to allow the user to select the desired number of simulations at preset levels, say 500, 1,000, and 2,000? I wouldn't be bothered by the extra processing time required.
Yes PRC runs 500 simulations. According to the article: ...simulations running 250 versus 100,000 scenarios varies only by about 1.5% for given levels of spending risk. However, the variation is wider at the extreme tails (0% and 100% risk), which provides some particular considerations for those who might be aiming for as close to 100% probability of success as possible. I think it would be resonabe for PRC to offer you the option of how many simulations to run but also provide the reasoning for more simulations concerning the extreme tails.
I think this has been discussed quite a bit over on Bogleheads forum. And in his excellent book, Living Off Your Money, Michael McClung provides a good explanation of his rationale for favoring backtesting using historic market results over Monte Carlo simulations, concluding:
Until a standardized model for markets is defined and agreed on, general Monte Carlo simulations are not attractive for estimating retirement risk.
Kitces resent blog called Gamification of Monte Carlo results is interesting:
In recent years, Monte Carlo simulation has become a popular tool for financial advisors to motivate their clients to follow recommendations. By presenting a single probability-of-success percentage, Monte Carlo analyses give clients a simple, instantaneous metric on the state of their financial plan. And because many clients naturally like to challenge themselves to do better and score higher, they are incentivized to take action that will increase their plan’s probability of success. The idea of using the same fun and appealing motivating elements found in games that people like to play (e.g., accomplishment, empowerment, and unpredictability) to encourage them to take action on other aspects of their lives is a concept known as “gamification”.
Yet, as many advisors know, the end goal of financial planning is not necessarily to achieve the highest possible Monte Carlo probability-of-success result, as a 100% Monte Carlo success rate effectively guarantees that the client will have excess money left over at the end of their lives (likely more than they would need to have at the end of their plan, and otherwise could have spent and enjoyed earlier in their life). Which means that, while Monte Carlo incentivizes clients to achieve higher and higher probabilities of success, actually working to achieve the ‘best’ success probability of 100% may push clients toward outcomes that are out of line with their goals for spending, giving, and leaving behind assets during their lifetimes.
Yep, when I'm reviewing Pralana results with my clients, if we see those solid 100% results I tell them right away they're not spending enough, and it's time to have more fun 🙂 That's a generalization, of course we dig in, make sure everything is correct, and strategize accordingly.
Early retirees or those just ready to retire do like the satisfaction of seeing the 100s though. That's not a problem, especially if one spouse/partner is more risk averse than the other, and certainly if both are. After they get into a groove and get settled in on their path while retired though, we want it dialed in more toward mid or maybe even low 90s, depending on their risk aversion profile.
If anything, it's always seemed to me that even >90% seems like overkill, especially if you are not planning for any heirs (much in the same way needing 80% of your normal income in retirement is ridiculous). 90% and higher always seems to leave a vary large residual amount on the table at end of life. I'd much prefer to adjust my plan annually as needed, give a lot to well-vetted charities as QCD's, and die with $1.75 in my pocket.
@nc-cpl yes, for sure. That's why it's so important to do this annually, and adjust that glide path.
Although I suspect that if everything stays relatively the same, it's just gong to keep giving you a 90% recommendation. Not sure how you'd use PRC to arrive at that annual spending amount unless you override the NSDS amount (go much higher), see what the resulting % success rate drops to, and decide (somehow) if that's an acceptable level of "depletion risk."
Another article from Kitces which appears to be a continuation of his latest one, this one is called - Reframing Monte Carlo Results To Increase Trust In Dynamic Retirement Spending: https://www.kitces.com/blog/monte-carlo-guardrails-probability-of-adjustment-success-client-communication-dynamic-retirement-spending/?utm_source=ActiveCampaign&utm_medium=email&utm_content=RSS%3AITEM%3ATITLE+++%5BNEV%5D&utm_campaign=NEV+Wednesday+Email
Ultimately, the key point is that outcomes, not probabilities, are what matter to clients, and any way of communicating Monte Carlo results should be clear about what those results mean in terms of real spending to the client.
@smatthews51 Kitces is putting on a webinar July 5th called: Improving Monte Carlo In Retirement Planning: Best Practices For Better Conversations
Derek Tharp, Lead Researcher at Kitces.com
Young and Pfau recently (1/10/23) published an article in Advisor Perspectives entitled "The Dangers of Monte Carlo Simulations":
The article provides an example of stress testing MC methods and the dependence of the outcome on Capital Market Assumptions (CMAs). The CMAs (returns, volatility, correlation) were obtained from a survey of "major investment advisory firms" by Horizon Actuarial Services and are used to produce pessimistic and optimistic success probability scenarios.
An example 60/40 portfolio test is provided in the article, and the Appendix shows the results for various stock/bond allocations. The article provides an interesting approach to applying MC methods and calibrating outcome expectations by stress testing based on CMA inputs.
The survey report itself (published in August 2022) provides further results for CMA values for 10 year and 20 year return horizons:
The survey uses geometric returns (annualized return over a multi-year period). A conversion formula from arithmetic returns (average return in any one year) to geometric returns is offered in the report. Note the discussion of inputting geometric returns in the Pralana Manual "Asset Classes and Asset Allocation" section.