Bogleheads® Live

Derek Tharp: spending in retirement

October 10, 2022 Derek Tharp Episode 24
Bogleheads® Live
Derek Tharp: spending in retirement
Show Notes Transcript Chapter Markers

[00:00:00] Jon: Bogleheads® Live is our ongoing Twitter space series where the do-it-yourself investor community asks their questions to financial experts -live on Twitter. You can ask your questions by joining us for the next Twitter space. 

[00:00:13] Get the dates and times for the next Bogleheads® Live by following the John C. Bogle Center for Financial Literacy on Twitter. That's @Bogleheads. For those that can't make the live events, episodes are recorded and turned into podcasts. This is that podcast. 

[00:00:30] Thank you to everyone who joined us for the 24th Bogleheads® Live. My name is Jon Luskin and I'm your host. Our guest for today is Derek Tharp. 

[00:00:40] Let's start by talking about the Bogleheads®, a community of investors who believe in keeping it simple, following a small number of tried-and-true investing principles.

[00:00:48] This episode of Bogleheads® Live as with all episodes is brought to you by the John C. Bogle Center for Financial Literacy, a 501(c)(3) non-profit organization dedicated to helping people make better financial decisions. Visit our newly redesigned website at to find valuable information and to make a tax-deductible donation.

[00:01:08] Mark your calendars for future episodes of Bogleheads® Live. Next week, we'll have Dan Egan returning to discuss systematic investment management and behavioral design. 

[00:01:18] Before we get started on today's show a disclaimer: This is for informational and entertainment purposes only. It should not be relied upon as a basis for investment tax or other financial planning decisions. 

[00:01:27] Let's get started on today's show with Derek Tharp.

[00:01:29] Derek Tharp is assistant professor of finance at the University of Southern Maine. He assists his financial planning clients at his firm Conscious Capital, and he is the lead researcher at, Michael Kitces’ blog Nerd's Eye View. He blogs on financial planning topics at, contributes regularly to the wealth management section of the Wall Street Journal’s expert blog. 

[00:01:52] Derek. Thanks for joining today on Bogleheads® Live. I shared your “Retirement Distribution Hatchet” article ahead of time on the Bogleheads® forums. For folks who haven't read that article, can you tell us what do Bogleheads® need to know about using risk-based guardrails to produce sustainable cash flows in retirement?

[00:02:12] Derek: Thanks for having me. Definitely honor and fun to be here, talking about risk-based guardrails today. 

[00:02:17] The first thing is to understand the different types of frameworks that are out there and all the assumptions behind them and how they're going to match up to reality.

[00:02:26] There's a lot of appeal to guardrails frameworks. Here I'm talking about Guyton-Klinger, Kitces’ ratcheting safe withdrawal rate. Some of these different frameworks that were very distribution rate driven. 

[00:02:38] It's based on a percentage distribution from the portfolio. and that's all fine if somebody's in a position where that's really how their income is going to be generated in retirement. Very consistent. They don't have cash flows changing. Say somebody's past age 70 they're already claiming Social Security, any pension incomes, not going to have any changes in rental income. 

[00:02:58] There's still some limitations. The biggest one being that the distribution rate that you can take should actually go up over time, as you approach retirement. Guyton-Klinger relaxes some rules over a certain point to kind of try and address that.

[00:03:12] But really, if you think about trying to keep risk at a more constant level, it actually should be a more of a steady glide path up as you live longer, your distribution rate, the likelihood you're going to continue to live X many years in the future is going down. So, you should be able to take a higher distribution from a portfolio. And that's usually missed in distribution rate guardrails frameworks.

[00:03:35] The whole idea with risk-based guardrails is to flip that around and say, “Okay, well, guardrails are great because they have some real advantages to helping us understand when we would make adjustments, how big those adjustments would be, how we kind of stay on track." But, flipping that to use something like a probability of success. It could be driven by Monte Carlo. It could be driven by historical simulation, but we could use probability of success as our risk metric. And what's nice about that is it's automatically going to capture changes with somebody's age, any unique cash flows, any unique spending goals that somebody has. That's all going to get captured when we have that metric instead of a distribution rate, but a probability of success. 

[00:04:17] So that's where we get to these risk-based guardrails. And then it's really the same type of framework. 

[00:04:22] Let's say somebody starts at a 90% probability of success. And they say if that probability of success goes up to 99%, now I'm going to spend more and maybe I'm going to adjust back down to an initial 90%. Or, if that probability of success falls, say it falls to 70% or 50% or wherever they put that lower guardrail, now I’m going to adjust back to try and stay on track.

[00:04:44] I think a lot of people, if they’re using a Monte Carlo tool, or if they’re using some sort of simulation tool for themselves are already doing that. But the guardrails take it one step further to say, “Okay, let’s put some concreteness to when these changes would actually occur and the size and magnitude of those changes so that we can be better informed as we’re going through retirement.”

[00:05:08] Jon: Derek, let talk a little bit about risk-based guardrails 101 and how folks can better understand the concept. With the traditional distribution strategies of 4% (or 3.3%, or whatever number you want to pick), you're assuming some future market return and volatility based on past performance of the markets. And you're also assuming a fixed life expectancy. 

[00:05:33] And then there's the Guyton model. So that's going to be, “Hey, the market is doing well, so I'm going to adjust my distributions accordingly." The risk-based guardrails approach looks at not just portfolio balance, but the fact that our life expectancy is now shorter. We should reflect that decreased risk, with respect to how much money we're taking out of our portfolio.

[00:05:55] Now we're adding this consideration of not just how much money I have, which is what the Guyton model does, which is a little bit more of an up-to-date approach, if you will, of just the fixed 4% rule or other fixed distribution rate. But with this guardrails approach, we're looking at not just portfolio balance ongoing, but also life expectancy and what that means for the risk to our retirement plan. 

[00:06:19] Derek: That's a good explanation of looking at those different ways of thinking about setting a retirement income level. The additional thing I would add to that is that with risk-based guardrails, you're really capturing all the differences.

[00:06:32] Life expectancy is one of them. You're adjusting that distribution rate. You're allowing for a steadily rising distribution rate, just as expected years remaining are decreasing. But you're also capturing other unique circumstances. A really big one that the whole concept behind the retirement distribution hatchet that we talked about was the blade of the hatchet.

[00:06:52] And the fact that when you look at distribution rates over retirement, let's say somebody's retiring at age 60, and they're trying to project out their distribution rates from their portfolio. If they're delaying Social Security to get that maximum increase in their benefit until age 70, what you'd actually see - if you mapped out how much they're taking from their portfolio - is you would have this kind blade of the hatchet where a distribution rates are higher in the early years cause they don't have Social Security to supplement their retirement income. And then once Social Security turns, then we get to the handle of the hatchet where the distribution rate declines significantly. 

[00:07:28] I’m just making up numbers. Somebody starts out at an 8% distribution rate, and then that distribution rate falls to 1%. That shape is very different.

[00:07:36] A very simplified example I like to use to think about the distribution guardrails framework is, let’s say you started at a 5% distribution rate. You say, “If that distribution grows to 6% of the portfolio, generally there meaning the portfolio’s gone down in value, then we’re going to cut back. Or, if that distribution rate falls to 4%, then that's going to be a sign that things have gone well. And we could spend more. So, maybe we adjust back to that 5%." 

[00:08:03] The problem there is, if you're trying to do that, and somebody is pre-Social Security, claiming their Social Security, or they have pension income or any other type of unique cash flow in the future, you're really not going to capture the unique risk profile that comes with that. And, there's things like it does heighten the sequence of returns risk. When somebody's talking about kind of front loading their distributions, taking more from a portfolio in the earlier years. The nice thing with the risk-based guardrails approach is not only do you get that longevity projection, steadily increasing distribution rate over time, but you're also capturing any of those unique cash flows, and things like the retirement distribution hatchet and capturing that change in distribution rates that very commonly occurs throughout retirement. 

[00:08:47] Jon: To say what you said a little bit differently, "Yes, we can consider that changing life expectancy over time, but we can also consider that the amount we're going to need from our portfolio is going to change over time. Once that Social Security kicks in, there's less of a demand for that portfolio. Our retirement planning should reflect that and using this risk-based guardrails approach, we can do that."

[00:09:12] Derek: That's absolutely correct. We really should ideally be accounting for that. Of course, there's other factors too. It could be that somebody knows a big retirement goal is they want to take several trips as a family and that's going to be a unique, large expense in their retirement. Any of those cash flows that we want to model in, we can do that in a standard Monte Carlo type framework. 

[00:09:35] The way I look at risk-based guardrails is it's capturing the best of both worlds. We get the analytical advantages of Monte Carlo. We get the communication and understanding advantages of a guardrails framework. We can bring those together and ideally have the best of both.

[00:09:55] Jon: Derek, we've talked about some theoretical numbers with respect to, "Hey, if your portfolio does this, then decrease spending by that." And you have similar frameworks in the article on for our podcast listeners. I'll link to that in show notes. 

[00:10:10] In your research, do you have some more definitive figures that folks should be thinking about with respect to, "Hey, how much should I be spending and how much should I be changing to my spending given a change in the value of my portfolio?" 

[00:10:24] Derek: An ongoing area of research and something I'd like to have something more concrete to point to. As a general note compared to the way that I think a lot of people approach and think about Monte Carlo simulation, there's a big difference between doing a one-time Monte Carlo simulation and doing ongoing Monte Carlo simulations through retirement.

[00:10:48] What I mean by that is, let's say you are beginning of retirement. You're going to run this Monte Carlo simulation once; you're going to set your spending level. And then, you're just going to charge forward blindly and spend that amount no matter what through retirement. I don't think that is an accurate reflection of how people actually go through retirement. You're going to continually monitor and adjust as you move throughout retirement.

[00:11:12] That distinction between the one-time Monte Carlo projection, set-it-and-forget-it, versus Monte Carlo as an ongoing spending and monitoring tool is not well appreciated. 

[00:11:25] I have another article on the Kitces blog, "Why 50% Probability Of Success Is Actually A Viable Monte Carlo Retirement Projection." When we did a similar type of analysis, we were looking at what if you use Monte Carlo was an ongoing tool and you planned for a constant 95% probability of success, a constant 50%, even down to a constant 20% probability of success. And what I think people would find very surprising is how small the differences in spending are when you use that.

[00:11:57] So much of your outcomes are driven by the actual market performance. What you're basically doing by using a lower or higher probability of success level is expressing some priority income for the current versus legacy. So, the higher the probability of success you use, the more legacy you're preserving, the less income you're taking now. You reduce the probability of success, you're taking more income now, but lower legacy values. 

[00:12:22] But in terms of even the maximum and minimum spending levels that historically somebody would've achieved if they went through retirement using those strategies are actually very similar and that it was quite surprising. 

[00:12:36] If you're going to be using on an ongoing basis (some of this planning), you really do have to think about probability of success differently. 50% probability of success is not as scary when you're using it as an ongoing tool as it is when you're doing it for a one-time plan.

[00:12:51] Jon: It is financial plannING. “I-N-G”. So, we're going to keep doing it. That's going to help improve our odds of success. 

[00:12:59] Adam: Just to play off, just that last part on the Monte Carlo, when you say using as an ongoing tool, how often do you think it should be used as an ongoing tool? 

[00:13:08] Derek: About once a year I'm comfortable with an update. If it's a situation where somebody's checking quarterly or semi-annually, more frequently, I don't think there's necessarily a lot of downsides to that. 

[00:13:21] When you look at intra-year volatility and how much things move around, if you get too responsive to something like this, you're going to catch a lot of false signals to make an adjustment when that may not actually be a true reflection of the path you're on. A once-a-year type update to a guardrail type plan is really plenty in my opinion. 

[00:13:43] Adam: And then my other more primary question is: if you think of someone who has 70% equities, 30% fixed income, and you're in a moment where the market is down like this, do you use these moments when you're taking distributions to automatically rebalance to the benchmark of say their benchmark 70/30, or do you try to maybe drive down a little bit more on the less volatile assets because that'll give time for things like equities, which usually take longer to come back to be able to compound and grow? So you don't end up knee capping a portfolio as it's recovering. 

[00:14:12] Derek: It partially depends on the strategy you're using, partially depends on how much you can be flexible. 

[00:14:19] Part of it’s even just the psychology of it. But, if you do have the ability to be flexible, I do personally like the idea of spending a little bit more from the less risky portion of a portfolio. Thinking of buckets and having something that's more of a growth bucket and more of a protection bucket and giving that growth bucket a little bit more time to recover, particularly when we're in a more pronounced downturn.

[00:14:43] We obviously must be comfortable with a changing risk profile of a portfolio over time. But if somebody also said, “I want keep things simple,” maybe they're using a target allocation fund or something like that, that wouldn't allow for that flexibility, I don't think it's the worst thing in the world either to go the other route. 

[00:15:00] Jon: Adam to comment on what you mentioned, “Hey, the market's down. I don't necessarily want to sell my stocks when they're down, so I'm going to sell some bonds instead.” Well, the catch there is you're taking a little bit more risk. 

[00:15:16] Now, arguably you have a little bit less risk because there already has been a market drawdown and given mean reversion, (which is a nerdy way of saying what goes down, goes back up and vice versa), arguably, you've already gone through the worst of it. But, folks need to be comfortable with taking more risk with that approach. 

[00:15:35] David: I look at the 4% withdrawal rule or the guardrail methodology. A lot of these withdrawal strategies are based on historical performance. Even a Monte Carlo simulation, it's based on what's happened in the past.

[00:15:53] Could we instead look forward more? Walk us through whether you think that type of approach would be more optimal rather than basing everything on what's happened in the past, like the 4% rule or something like that. 

[00:16:07] Derek: If we're using historical as the foundation, then by definition yes, absolutely, it's purely backward looking.

[00:16:14] With Monte Carlo, it obviously comes down to what we're using for our capital market assumptions. It is very common that those are historical averages. 

[00:16:24] A couple different schools of thought. One would be that they're just purely historical. And then yes, I think the same would apply. There's another approach to use forward-looking capital market assumptions, which are trying to do something similar in a different way, to what you're describing here, where we're saying, 

"Returns may not actually be as nice as they work historically, given the current environment we're in. So we might project something out, that's a little more pessimistic."

[00:16:51] You must be careful with that because that can easily steer you too conservative. Using these reduced return assumptions and projecting all the way out for 30, 40 years, to me that might be overly pessimistic. My preferred way to try and capture some of the current environment and factor it in is a ‘regime-based’ Monte Carlo.

[00:17:16] We might one set of return assumptions for a 10-year period. And then once we get out beyond that, then we drift back towards longer term averages. It is about that 10-year time period where we have the most historical reliability in terms of being able to make some adjustments. When you get out beyond 20, 30, 40 years, and then things don't necessarily have as much predictive power as you would have at that 10-year mark. 

[00:17:41] David: Could you not argue that, especially for bonds, when you see yields go up, as they have over the last six months or even equity prices when they go down, expected returns mathematically are higher. 

[00:17:53] So this idea of spending less when markets are soft, especially when bond prices are down, could that not lead to kind of suboptimal behavior because expected returns are higher?

[00:18:09] Derek: The risk is always that you're reducing spending some (and I personally prefer spending reductions that aren't going to be drastic), but you're reducing spending some to guard against the possibility that maybe you're headed down a path of a prolonged, lower-than-historical-average period. 

[00:18:28] When we look at the safe withdrawal rate research, we're usually talking about just the sequence kind of mid-1960s that drives the 4% rule. It's that one historical worst scenario based on both returns and inflation and everything that happened over that 30-year retirement period.

[00:18:45] What you're guarding against, is not so much that long term average expectation, but what if you're headed down a particularly bad sequence of returns. 

[00:18:53] But to your point, I do think you're right that as markets decline, capital market assumptions should also probably be changing. And what should be happening using a regime-based Monte Carlo approach, is that now as the market has moved really we should be changing those forward looking capital market assumptions.

[00:19:14] The net effect of that is that it should diminish or reduce some of the impact. Saying, 

“Yes, Okay portfolio values are lower now, but now maybe we're actually using higher expected returns for the next 10-year period that we're projecting out.”

[00:19:29] It could help avoid an overreaction than you might use if you're just using static capital market assumptions and not reflecting that the economic environment that you're in has changed. 

[00:19:42] Jon: Here's a question from the Bogleheads® Forums from username IIM7V7IM7, who writes: 

[00:19:50] The financial planning industry tends to work with a client to design a long term, 30 year or longer, plan based on their own understanding of client goals, assets, cash flows, and liabilities. In other industries that have long term multi-year developments, they create milestones where a plan is reviewed and adjusted for a new forecast accordingly.

[00:20:08] Why doesn't the financial planning community admit that there are simply too many variables to accurately forecast a 30-year plus plan with any degree of accuracy? Start with an initial plan and set up a development project and milestones to meet and reassess such as Medicare costs, Social Security RMD, halfway points, such as 15 and 30 years or any major unplanned spending shock to meet and assess the initial plan? 

[00:20:30] And Derek, correct me if I'm wrong, I feel like this is precisely what the risk-based guardrails approach does. Look at this annually. Figure out if we're still on track and if not, we'll make some adjustments to our spending. 

[00:20:42] Derek: Yeah, I do think so, and to go back to the statement about how it’s impossible to do any type of forecasting. What's the alternative? You must do something in terms of trying to make some sort of projection or have some sort of framework or model.

[00:20:56] It really should be more about these milestones, if you want to call them that, or check-in points. For good ongoing planning. The, I-N-G. Like we've talked about, that to me is an ongoing process. And I think there are a lot of advisors that do that, looking at all these things, revisiting them, adjusting for anything that we know. 

[00:21:13] This all fits into the risk-based guardrails as well. We might have started out saying a client wants to spend a certain level in retirement, or they have certain goals that they're planning for. Maybe it's those family trips in the future. And maybe a year or two passes, and they're doing a plan update and they've decided, you know what, now they actually want to double the size of the trip, or they want to give more to an organization or whatever it might be.

[00:21:37] As the plan adjusts, where the goals have changed, where the risk profiles changed, health has changed, those are all things that should on an ongoing basis be reflected within the overall plan. And as long as they are, then that should come out in the risk-based guardrails. 

[00:21:54] Jon: We've talked about using Monte Carlo tools to help us implement this risk-based guardrails approach. For the do-it-yourself investors that make up the Bogleheads® community, what are some options for taking advantage of Monte Carlo simulations?

[00:22:10] Derek: In the do-it-yourself type of realm, there's varied abilities. I've seen some people that have built incredible software and their own tools and their own modeling. It's a hard question to answer from that perspective. In terms of what somebody could build themselves, it can really vary. 

[00:22:25] Let's say you're in a situation where you don't want to build out a tool yourself, you're trying to look for something to use. The part of the risk-based guardrails that's really the trickiest, is the part that you may not have to use. It's communicating the guardrails; where would a portfolio go to before you had to make an adjustment? How big would the adjustment be when you got there? 

[00:22:46] That is a part of the guardrails that, if you wanted to take that piece out and you just wanted to use the Monte Carlo tool and you just wanted to come up with a policy that said, (throwing out numbers here for illustration purposes), 

“I'm going to start spending at 90% probability of success. If it goes to 99%, I'm going to adjust back to 90% probability of success. Or if it falls to 50%, I'm going to adjust back to wherever my new target would be.” 

If you wanted to put something simple in place like that, you can just watch the probability of success move over time and then adjust accordingly.

[00:23:21] It's very hard to get an understanding of how much spending volatility is involved with a certain strategy. How much ups or downs you must tolerate or be willing to go through. So that's the trickier part, is understanding the strategy of the trajectory you're on. So, if you have access to a basic Monte Carlo tool, that's a way to go. 

[00:23:43] Jon: Derek, any final thoughts before I let you go? 

[00:23:47] Derek: It was a lot of fun to be here and it's a different type of concept we're covering here today, but if anybody has questions feel free to reach out. I'm always happy to chat and think about different areas for future research as well. So always appreciate that feedback. 

[00:24:02] Jon: Folks, that’s going to be it for all the time that we have for today. Thank you to everyone who joined us for today and thank you to Derek. 

[00:24:08] Next week, we'll have Dan Egan returning to discuss systematic investment management and behavioral design. 

[00:24:15] Check out a wealth of information for do-it-yourself investors at the John C. Bogle Center for Financial Literacy at

[00:24:23] Finally, I would be honored for some feedback. If you have a comment or guest suggestion, tag your host at @JonLuskin on Twitter. A big thank you to Ryan Barrett who put together episode summaries of episodes 5 and episode 19, and another big thank you to Richard Feldman who put together the transcript for episode 16.

[00:24:41] Thank you again, everyone. I look forward to seeing you all again next week. Until then, have a great week.