Episode 407: Michael Kothakota - The Shape of Financial Planning

21:31His background's nuclear engineering, very similar to mechanical engineering framework. He and I talked deeply about that. And so I ended up spending a lot of time with him. And so Richard Bellman, I don't know if he's the founder of dynamic programming, but it's basically trying to break these complex sequential tasks into a simpler subproblem. In my paper, the extension or the HJB equation or the Hamilton-Jacobe-Bellman equation is designed to characterize optimal behavior. When we talk about how this can possibly work at scale, we'll get into why that is so challenging. because having the formula is one thing.

22:11Deriving it for an actual household is extremely complex and very difficult to do. The big thing dealing with dynamic programming is that in this particular case, we have these value functions that are non-smooth. There are discontinuities in different parts of the plan. Some planning strategies don't exist until you reach a certain age. Sometimes they're gone after a certain age. Sometimes they don't exist in certain countries, for example, or if you're an expat. So there are a number of things that create these kinks. Sometimes people write in their papers, they call it kinky. So there are kinks at these transition points, basically.

22:50The idea is that the value function is basically to say, okay, if we track all of these things, and we optimize for this particular case with these people's particular values, then that's better than not integrating. That's better than somebody coming along and saying, hey, I'm going to build a retirement plan for you. And then you go to your tax advisor and they file your taxes and they do your tax planning. It's better than that siloed approach because we're considering all of those different things in order to make it optimal.

23:23Ben Felix:This theory kind of explains why integrative financial planning is valuable, which is interesting. You're able to sort of quantify why that advice is valuable. I have a question. Empirically, is there evidence that the financial planning profession actually offers measurable benefits to households? Sort of. I think the challenge is in the data sets. So Sherman Hanna had an article a number of years ago. I talk about it in the paper, but basically we use a lot of these in graduate school. In social sciences, we use a lot of these large data sets. other than probably in the last few years, none of them are designed for actual financial planning.

24:03They're household finance. They're generally designed for decision-making. They're designed for economists to kind of make decisions about and observe changes in household dynamics and then make some policy recommendations. And so that paper, it will identify, okay, these people have a financial planner. Are they better off? And it'll make some comparisons and make some assumptions. And so you'll see a lift based on that. So the idea is that if people who are advised are going to have better financial outcomes than people who don't. And then more recently, so the project that I started at CFP board was the longitudinal study.

24:43So basically the idea was, okay, well, let's look at financial planning outcomes. Let's do this longitudinally. The questions are designed to be more financial planning specific and to deal with like financial wellbeing. so it pulls in the objective pieces and then also pulls in some of those more mental well-being, psychological well-being pieces so that it is a little bit easier to identify. And so that's what Dr. Heckman, Dr. Luter, Dr. Cushel, and what's Michael's last name? A guy from Wisconsin. I'm blanking on his name right now, but he's brilliant as well. So they've been working on this project.

25:17I think they're in their third year of data collection. They've been able to see both in the first year a lift in people who are advised versus not advised across almost every dimension. So there is some empirical evidence there. I think where you run into things is like, well, what's the quality of that? What are they actually doing? So okay, great. We know that they're doing it. But what are they actually doing to provide this service? Maybe they're just there and the stuff I'm talking about doesn't matter. Or maybe they're saying, okay, look, I'm really going to take a look at how everything is going to affect you.

25:51But because we don't know that and it's really difficult to get at, we need a different set of research designs to actually get at that information, which something like that large scale is not designed to do and nor would it do. And then I think sworn strategy had pretty similar results. But again, it was one of these big data sets that is less useful.

26:12Ben Felix:There's always that selection bias issue to it in the large studies like that because financial planners are going to be more likely to seek out clients that have better financial situations because they're more profitable to serve. Right. Are they coachable or they have less mental health barriers? Absolutely. And I think it's interesting. So when we drafted the initial FPLS survey, Dr. Hackman spent probably, I want to say, months on trying to figure out selection problem. And so it's about as good as it can be. Dr. Hackman was actually my dissertation advisor. It's about as good as it can be from a selection perspective, but you're absolutely right.

26:48You're just never going to get all that eliminated.

26:51Cameron Passmore:How would you describe the integrated financial planning architecture that underlies your theory? The first piece is that structure, which is the structural tensor. So tensor can be an abstract principle, but basically kind of think about it as like a multi-dimensional object. Financial planning has got a lot of dimensions to it. So it's just a way of capturing that. We know that there are all of these dimensions. Part of the financial planning problem is to how can we reduce that dimensionality to make it solvable? Because if it's too complex, nobody can solve it. Those structures exist regardless of what clients do or feel.

27:28I mean, obviously, they can enhance their financial position using that structure. But in general, it's things like tax law and stuff, I think. And then there's the priority piece. And so that's how do people prioritize so that all those weights sum to one. One of the things that's interesting about this that I've been playing around with is I did use the CFP domains. If you think about it, that's kind of arbitrary. These are some domains that you could actually split it up a number of different ways. The Ultra High Net Worth Institute has 10 domains. You could look at it that way. There's a variety.

27:59If you've included businesses, where does that flow? Is that an investment? Is it part of retirement? it, there are different pieces. And depending on how you define the objective functions, you could change a lot of this. The idea was to make it a little bit... I said I was leaning into the complexity, but at some point, you can't. It becomes unwieldy. And then you have the discount matrix, which interacts directly with both the structural tensor and then the priority weights. And the idea there is that you then get this discount-adjusted coupling matrix where you can assess the strength of the coupling between the domains based on how somebody values them at a certain point in time.

28:38It's just a way to kind of assess how the planning strategies can affect somebody's financial picture. And so in order to do that, you have to build a structure that can support that kind of thinking.

28:53Ben Felix:You talk about this in the paper, but the human financial planner is kind of this like processing machine that thinks about all this stuff sort of subjectively. and we do it like we will go through a financial planning projection and show like, well, if you do this, this is the outcome. If you do that, this is the outcome. And then we'll kind of get a feel for how the client responds to that. But you've quantified that, which I think is really cool. And I want to kind of talk about like where my thinking of that came from is that something as simple as if I throw a baseball to you and you catch it, your brain did a bunch of stuff.

29:23A lot of it was unconscious, but essentially it did some calculus. It figured out what the trajectory was. Actually, I did talk to somebody who disagrees with me, who's a mathematician. But I think that you basically applied this math to it because you figured out how to catch it. You've identified where it's going. You put your hand out in time. You squeezed your hand in time. That complex thing was distilled very quickly. You didn't have to write out an equation on how you were going to catch it. I think it's very similar in this respect. So if you're an experienced advisor who's trained, you start to kind of understand how to kind of reduce those dimensions naturally.

30:01And yes, it is a little subjective, but based upon some experience.

30:07Ben Felix:It's intuition gained from professional experience. You mentioned the tensors. I learned a bunch of words from your paper that I hadn't really seen in this context before. So your paper also talks about curved surfaces, cornered spaces, binding constraints, manifolds and corners. I'm reading this like, what is happening? Can you talk about what that stuff means? What is the shape of financial planning? And why does that matter to the understanding the overall system? So I'll be the first to say it's weird. I thought about this as like spheres. And then I thought about it as a series of interconnected spheres.

30:38And then as you start to think about it, there are a lot of constraints in financial planning or most constructs. So a sphere is not infinitely expandable debt, they're borrowing limits. And so if you have all of these different interconnected things, which at first I thought, well, that was how I was going to solve it with just like topological spheres. By the way, I thought that that was what I thought the breakthrough was. And then I started running into all these problems. And so what a manifold is, and it actually doesn't have to be a manifold. Manifold is just, you can think of it as like the financial state.

31:09If we go back to the wave function, it's like the financial state at any given point in time and there are different probabilistic outcomes to that financial state based on decisions that are made. So that is what the manifold is. I think there's this guy at... What's the name of that university? It's in Kansas, not Kansas State. This guy, he studies ring theory. And so if you know what a Taurus is, which I didn't know what a Taurus was either, but a Taurus is basically like a donut, a ring, essentially. This theory can work with in that if you can imagine that space. But I also felt like that even that just seemed pretty hard.

31:44So the manifold is just like, we're just going to say, hey, look, this is the state space. And this is your financial state and the probabilistic structure of a client's financial state, their life. And so a lot of the constraints will bind at the same time. And so that will produce these corners. You can't have negative wealth in any account. If you have negative wealth, it's basically debt. Assets, really, they can only go to zero. So you have that. You have borrowing limits that are tied to income. You have liquidity requirements that are for upcoming expenses. You got a number of issues. And so what happens when two or three of those will bind at the same time, when they occur at the same time, given the set of circumstances?

32:27So if you reach your borrowing limits, you don't have enough liquidity for your expenses, then you've got what we would call a corner. Why does that matter? Who cares if you've got a corner? If you're trying to move around in this room, we've got a lot of corners in this room, but if you're moving around a room and the walls are curved, then you've got three or four walls meeting together. Let's say like a T or something like that. You can't move around that wall specifically because it's bound in multiple places. I think that's kind of the easiest way to think about it. It's a little bit more complex than that.

33:03That's why that matters. But then why does it matter for this paper? Standard optimization assumes smoothness. Smooth, unconstrained, which makes it easier. And it makes it a much easier kind of problem to solve. And so a lot of the optimization that we do and a lot of the optimization that operations research and decision scientists do, they start with smoothness. But we've got these kinks because of how financial planning works. We live in a non-smooth world. And so we've got to respect that with whatever math we're going to use to describe it.

33:36Ben Felix:Like simple two-dimensional calculus, you can fairly easily do an optimization. But you're talking about doing optimization over a three-dimensional surface that is not smooth. I mean, that's what you're trying to solve. Exactly. It's not differentiable at that corner. That's actually a way better way of saying it.

33:53Cameron Passmore:So how do the fiber bundles come into your theory? So if the state space is the manifold of people's financial states, well, we as financial planners or even individuals, we execute strategies to optimize our financial life. In most utility functions, we wouldn't typically have things that change. The fiber bundles are your strategy space. And it's just a way to think about that the strategy affects the manifold. And so the way it's visualized in the paper is that the fiber bundles sit above the state space, and then they're going to change. So your strategy space changes for a variety of reasons.

34:29Here we have Medicare age. Things you do with Medicare when that occurs are different than what you would do before. Social Security exists for the time being here. And at some point, you can draw on that. Even though you die, a lot of times, a lot of estate planning is, well, what happens after you die? There are a lot of strategies that can occur up to that point, but then less that can occur after or maybe even more that can occur after. So they change. And so fiber bundles, they're a way of thinking about how the strategy space evolves over time as well. In addition to those structures, those things that happen, a lot of other things can happen.

35:10Legislative changes. One of the things I do with clients is we do an annual simulation where we'll simulate a macro event and then we'll start to also simulate some smaller level things that are idiosyncratic to them that could affect their lives. I don't use this one because it's pretty catastrophic, but if Yellowstone erupts, that's a pretty big deal. That's going to affect a lot of things. That changes everything. The collapse of the dollar, the current decoupling and transactional relationship that my government is currently engaged in, that has changed how we interact across borders. That changes a lot of things.

35:47And so that's what those fiber bundles are supposed to represent, is to see at these different phase transitions and then different events that occur, the strategy space changes and we need to adapt to those changes. That's how it captures it.

36:03Ben Felix:We've got this kinky three-dimensional space that's all weird and stuff, and then we're trying to solve it with strategies that exist. And so you can do that optimization at a point in time, but then as we move through time, the available strategies are going to change and that's going to change the optimization process. That's essentially what's happening is that your strategy just change. And what you do for a client at a particular time, divorce, that's another thing, death of a spouse, the strategy set changes. This is actually my counter argument when people say it's a little too complex is that we don't have access to the same strategies or need of the same strategies throughout the life cycle.

36:41We need to have the ability to adapt and that's the point of the fiber bundles.

36:45Ben Felix:In the mathematical model, what domain objective functions are being optimized over? So cash flow, that's a very simple one. It's margin between income and expenses, liquidity reserves, debt service coverage, retirement security in this case, this particular one is maintaining a target income through life expectancy. You could use replacement ratios, you could use shortfall probabilities, I actually modeled a number of different ones in some cases where the numbers can be higher, frankly, depending on what objective function you use. Tax efficiency, so that's tax rate. Current and future, that's a little bit hard to model because although I always feel like in this country, like, okay, well, they're going to get lowered every four to eight years and then they're going to get raised every four to eight years, just in general.

37:34So I just modeled them as based on income and whatever the current rate is. estate effectiveness. So that's how effectively were assets transferred to heirs versus charities and then align it with whatever their legacy intentions were. Actually, Ultra Net Worth Institute has this really interesting function that they developed for how that works. Risk management adequacy just deals with coverage, making sure that the risks that we typically manage can be managed, although that can be adjusted too. And then across countries, I think there's a lot of fascinating ways to do that. The investment performance is just a risk-adjusted return using like Sharpe Ratio.

38:15But even that, you can use any sort of model to include in there. And I did test a few others, but basically very similar results. The challenge is like how you can actually simulate this across all domains. It's really difficult. Those are the objective functions that kind of flow. They become unitless. they result in the effect of tensor and then they're recalculated out in the value function. Wild.

38:39Cameron Passmore:In your paper, you assigned a weighting function to each of the domains like cash flow, weighting on cash flow, tax, retirement, and so on. But would you think about assigning a weighting factor to each objective? Would you potentially change the objective function depending on client preferences or would you just leave it at the level of assigning weights to each domain? I was presenting this at Texas Tech. Nobody asked the question, but I did mention it. you could argue that it's arbitrary. I chose these functions. A, I didn't think of the idea of weighting them. I like it. I also didn't think of this objective function as best for a particular client.

39:15I like that too. What I did think about was that I want researchers to say, okay, well, let's test these. Let's go through and let's actually test to see on real situations if we can, what is the best objective function? But my guess is it's very similar to like the priority matrix and that it's idiosyncratic. That's a really good idea. I like both of those. The point is like for people to take this, test it, see what's working, what's not. Maybe things need to be adjusted. Maybe it's not actually representing what I think it represents, but I like that.

39:52Cameron Passmore:We talked about the natural kinkiness of financial planning. It's very kinky. So how do you solve for that when the objective functions aren't smooth and the topology is not smooth? but how do you actually work around that from a mathematical perspective? There are two ways. One was to deal with the non-smoothness used what we call viscosity solutions. So you guys think about it as if it's non-smooth, if you have like a thick liquid over a corner, that's the way I kind of visualize it. You kind of put it over some honey over a corner. It's kind of smooth at that point. And then there's Clark generalized calculus.

40:29to their integral differential equations. So I had to spend a lot of time learning. It's a combination of integrals and derivatives for folks who don't know what those are. But the Clark generalized calculus kind of helps with handling the kinks. It's a way to capture a set of all the possible limiting directions. So if you kind of think about, you know, if you've got your function that's moving, it could kink up or could kink down. it's a way of capturing like those potential changes. In most cases, things like a tax bracket is kind of easy because you know, it's probably going to go up. But I mean, if you're talking about a state or something like that, and you're trying to optimize in the US anyway, we have very low thresholds for like higher estate tax.

41:17If you're trying to optimize between how much for the trust to pay versus how much for the beneficiary to pay, there's a lot of interesting things that can happen there. That's essentially the two methods that I used. I think you could actually use viscosity solutions on both. I think I tried that. And then what happened is you lose the assumptions of uniqueness and existence in the Hamilton-Jacobi-Bellman equation.

41:48Ben Felix:Viscosity solutions in financial planning is not something that I would have... But it's a word though that makes sense. I get it. Yeah, yeah, yeah. Clark Generalized Calculus is just like a guy who had it, you know, it's named after him. So that's a little bit more challenging. But like, if you know what integrals and derivatives are, and they combine and how you can deal with those different directions, I think financial planners should be able to be like, okay, I get that. I get what that means. Oh, yeah, it makes sense. Once you've explained it all. If someone told me that we're gonna be talking about viscosity in a financial planning conversation, I would be like, what are you talking about?

42:18I think financial planning is like the stepchild of research. And so every other field has all these really cool things that if we just look to those other fields, we can pull them in and then they start to make sense. Can you talk more about how the priority matrix in the model works? Six by six matrix, diagonal matrix. There are weights for each of the domains, each of the planning domains. And so it's designed to capture what's important to a particular client. the idea being in general, this is how somebody will prioritize these things. And so the big deal from my perspective is, I think a lot of the things that we say are irrational behaviors are just preferences.

43:01There certainly are things that are irrational. Originally in economics, when we talked about people behaving rationally, it was consistent. It was behavior that was consistent with how they wanted to live their life. And then in order to make things work, we were like, no, these are the things that are rational. wealth maximization, all of these things. These are the things that are rational. So if you deviate from that, then that's irrational as opposed to, well, wait a minute, people have these different preferences, which of course we attempted to capture in a variety of things, but in economics generally, like in difference curves and things like that.

43:34So each one of those categories is weighted. So you ask people like, how important is this to you? And there are a variety of ways to do this. Financial planners do some of these things naturally. And some of the questions to make it work in the paper, like, hey, rank order these. That's hard for people to conceptualize. Well, what do you mean rank order them? And then now you've got to still move the weights around to make sure that it works. Revealed preference tests are a possibility. You could do like some pairwise comparison of how people think about the different ones. And so then essentially what you do is you end up with a set of weights that then get applied to each domain.

44:14And then when the structural tensor gets filtered through that, then you now have something that's more appropriate for that client. And originally, there was no discount matrix in this. It was basically like, okay, well, yeah, this makes sense. But then I started thinking, things change over time and how that works.

44:34Cameron Passmore:Can you dig into that a little bit more? How does the relative urgency of various schools get reflected in the model? So in my class, I teach a weird combination of financial mathematics and ethics. One of the things that I have students do is, especially in the age of AI, is like if they're solving like a simple time value of money problem, I actually have them create the discount rate. And discount rate construction in a normal problem, a CFP exam, CFA exam, usually given like a discount rate. Business valuations, sometimes you're given a bunch of things that you create a buildup rate, you'll create a discount rate through a variety of approaches.

45:12Intuitively, what we do is we apply those discount rates to different priorities. So something like retirement, maybe I'm putting money in a 401k at 25, or I'm thinking a little bit about retirement, but it's not urgent. At 50, and I realized I've saved$50 ,000. Well, it starts to become a little bit more urgent. I started thinking a little bit more about it. And so the discount rate, it gets applied to that priority. It's like, okay, it's important, but it's not important right now. The higher the discount rate is typically the more urgent it is. Risk management has a half-life of 10 months because you need coverage now.

45:49You need to make sure that if you get a car wreck, that you're able to buy a new car, you're able to pay your medical bills, etc. But estate planning, typically that might be really far off. I started thinking about that a little bit more too. I'm like, but are those fixed? I don't think they are. So they can be treated as random because there are a number of things that happen in your life that change, that change you and change how you think about stuff. I started out in the army and now I'm a financial planner talking about tensor mathematics, fiber bundles and manifolds with quarters. You've got like that baseline, which would be like the long run mean, like basically how you think about that particular thing.

46:29and then you've got those endogenous adjustments. So as your financial state changes, you start to think and your age changes, you think more about different things. And then you've got the exogenous shocks, the divorce, the death of a spouse, the birth of a child and all of those things now change how you think about stuff. And so that will affect those rates. So they're treated stochastically in the paper because of those things because we actually don't know how all those things are going to play out.

46:59Ben Felix:The discount rate for estate planning will be one thing now, like when you're sitting here recording a podcast is one thing. But if you have a near-death experience or if a spouse dies or if you get some kind of diagnosis, the discount rate on that could change. Is that kind of what you mean? Yes, all of those. And then they see what their parents had to go through and they had to settle the estate and they're like, wait a minute, I don't want my kids or my wife to have to do any of that. So now all of a sudden, it's much more of a priority. We need to get these things fixed. And so something like if you get a health diagnosis, how that's going to affect all of the different domains, right?

47:34So it's going to increase urgency in a lot of things. It's going to increase urgency in estate planning, your risk management and your cash flow because you're like, well, wait, I got all these bills. Maybe I can't work and I need to make sure that my family is taken care of.

47:46Ben Felix:The reality is that can change over time. And so when you're building an optimization, like, okay, how do we build an optimal financial plan? What does that look like? You have to account for the stochastic nature of that variable. You have to try to. I mean, I think I went back and forth on this a lot because I thought, well, the wave function, like we just don't know what that state's going to be. Estimating is the best that we can do. I think we should treat them stochastically because they are going to change and we don't know ahead of time. But people do think they know how they feel at that moment about each of those priorities.

48:21They can wait them then and they can think about how urgent is it. They can think about that now. But then when things change, those things change. So we can both take their word for it and not take their word for it

48:32Ben Felix:because of all those things that can occur. Should financial planners be influencing those discount rates? I think they should explain them. I think they should be talked about. I think it's a useful exercise to talk to clients about these things, how they weight things. A good example is because people use risk tolerance all the time. And so sometimes it doesn't matter what somebody's risk tolerance is. Maybe they have a really high risk tolerance, but they've already got$400 million. Well, who cares? They can do whatever they want. But if somebody's got a really low risk tolerance, but they're not going to be able to meet any of their objectives, they need to be educated.

49:07And so I think it's similar in this sense. We could spend a whole other podcast on risk tolerance. And my thoughts on that. It's similar here where it's like, you're not prioritizing retirement, but you will. You're 30, you're unmarried, you love your job. So you're not thinking about it, but like, here are all the things that can occur. You need to think about how you would prioritize that. Same thing with like insurance. People don't get life insurance because they think nothing can happen to them. Well, let's actually look at, first of all, low probability events happen all the time. It can happen to you.

49:39Is it likely? No, but like, what are the costs? associated with that. And then you could dial in like a discount rate. I think if you can show people like what that means mathematically, I think it's a useful exercise.

49:52Ben Felix:To re-anchor listeners, we're talking about like how important the different financial planning domains are to someone who is making financial decisions today. And someone who's very far away from retirement might not think that retirement is very important to think about, but we're saying that the financial planner in that situation should maybe be talking to them about why they are going to care about that one day, which should influence how much they care about it today. Because with compounding and stuff, you can't solve for it later the same way you can solve for it now. The way I would approach it with a young client is to say, I get it.

50:23I understand why you're not prioritizing that. Absolutely. I want to walk you through some things. Even if you're a new planner, you could say things like, I've seen cases where XYZ has occurred, and then all of a sudden, it's too late to think about retirement. Or you now have to sublimate a different value. People talk about like work-life balance and flexibility, temporal flexibility and stuff like that. Well, if that's a priority for you, now you can't do it because you're 50 and you didn't save for retirement. So let's think about it. That's important to you. So like, let's talk about that.

50:57I think it's absolutely something we should be talking to people about.

51:01Ben Felix:Super interesting to think about. Can you describe the four-tier architecture that summarizes your model? The structural tensor is, It's a rank three tensor. So basically you can think of that as like a three-dimensional object. And then it's basically, what it's doing is it's, you've got different domains and they affect each other. A affects B, which affects C, which affects D, but A also affects C, which affects D, which affects E and so on and so forth. So there's a lot of different interconnectedness. If you do something in one area, it's going to affect all these other things to a certain extent.

51:36So some of the couplings are strong. some of them aren't. One thing doesn't always affect something else. That is the objective mechanics. And then what matters most to the client, which is the priority weights. We get those through discovery meetings. We get those through learning about the client. I don't think you really know a client completely for a few years, at least. And then there's the discount matrix. That's the urgency structure. So how urgent is something to you. And then those get synthesized together, which is the complete change from the objective structure, the subjective priorities, and then the urgency.

52:13The practical purpose is that theoretically, we can pre-compute the structural tensor. We can update that as regulations change or social dynamics change or whatever. We can update those things. And then the priority matrix, we have to get that from clients. That part makes sense to separate. The other thing it does is it reduces computational burden. Because if you try to do all that at once, it's a lot of parameters. It's more personal if you do it that way. That's the architecture in short.

52:43Cameron Passmore:So how are you capturing uncertainty in your model? Some piece of it gets captured in the discount matrix as far as how people feel and how people think about each of the domains. We have to think about, so there are the things that we can plan for that we do. Okay, you need X insurance adequacy. And so we can treat that uncertainty with either self-insurance or risk sharing or risk transferring to something else. In investments, for example, investment volatility, that's exogenous. But like portfolio exposure, it's in Dodge. It's how you created it. It's within the system. You created the exposure.

53:21Longevity risk is exogenous, but the financial consequence of longevity risk is endogenous. You've got the endogenous uncertainty, and then you've got exogenous uncertainty, things that are outside of the client's control, but we could anticipate. Things like tax law changes, macroeconomic conditions, regulatory changes, pandemic. So we can anticipate that, but we don't always do it. Before I even finish this, this was a piece that I thought about really deeply. and also was unsure how it would actually fit into the theory. I do this with clients. So I actually talk about a lot of these exogenous shocks.

53:59And what is interesting from the perspective of the client is that just planning for it, something that may never happen. The year before simulation was China invades Taiwan. I usually try to pick something that's not too on the nose, but that could happen. We looked at the cascading effects of how that would affect them. And just going through that process is really helpful in showing that you care and that you're thinking about things. But also, it creates a little bit more connection, involves them in the process, makes them feel a little bit more in control of a lot of these things that are actually outside of their control.

54:34It's not something they can do about. We can only react to those things. That's how we capture exogenous uncertainty. There's the priority uncertainties. Priorities shift as life unfolds. and then there's the rate uncertainty with the discount rate. Those are the four ways that I try to capture uncertainty in the model.

54:54Ben Felix:I love that point about we don't do specific simulations like what you're talking about, like event simulations, but we do basic Monte Carlo type stuff. And it's such a powerful thing to be able to say to a client when the market's down 10 % or whatever, to be able to say, well, we factored this type of outcome into your financial plan and it's still good. That's why we did it this way. I think that's always reassuring. But the idea of doing a specific event is a pretty cool idea. I don't have hundreds of clients. So I try to keep it small on purpose. So like it is time consuming. Somebody with 500 clients should not do it.

55:26A hundred or less, it's probably something that's doable and they love it. It's kind of a fun thing

55:30Ben Felix:to do. You built this wild model that we've been talking about. What are the main theoretical results from the model? Somebody actually mentioned this the other day, saying it was empirical proof. It's not empirical proof. This is theory. It's using client fact patterns that exist for me. And then anonymizing a bunch of it. It's very theorem heavy. And I felt like that was necessary to show what was happening. So there's the structural piece of it. That structural coupling really only matters if a client assigns priority to it. So that's one. There's the stability and separability. So that's essentially like if you have a systemic event, the stochastic separability conditions are more strict than a deterministic one.

56:17Time value of money projections are very deterministic, but obviously don't capture a lot of things. When we treat those domains independently, it can destroy that separability. There's an integration premium. So there's the value that you provide in each of those domains. So like as a financial planner, Even if you're not fully thinking about the interdependent nature, you're still providing value. So like if you think about things like Morningstar Alpha or Morningstar Gamma, Vanguard Alpha, we use all the Greek stuff here. Priority Lambda is what we have here. There is a appreciable difference from the actual integration.

56:55If the separability conditions are not met, what ends up happening is the integration premium goes away. And so it doesn't actually matter. That really doesn't even make sense, but it will never be less than zero. So the integration premium will never be less than zero if it's done correctly. Dealing with the jump rates for the uncertainty events, if you're accounting for things like that, that can boost the premium even more. It would be great to be able to see this happen in the wild. Those are the main theoretical results. Although in the paper, I think I mentioned this at the beginning, I also added a value loss.

57:29If you don't have a good idea of what the client's values are, if it's misaligned, It's like a mean 4.8 % loss in value if you don't actually identify the client's priorities well, which doesn't sound like a lot. It can be quite a bit depending on how long that lasts. And I'm not saying that we always will have perfect alignment on priorities. That's one of the reasons it's a big deal. That's one of the reasons I think that human advisors matter is because it's very difficult to do. They change. There are some significant costs to that.

58:02Cameron Passmore:So how should financial planners think about applying your mathematical model to real financial planning scenarios? The first thing is to really think through how a course of action they might be thinking of recommending or their firm is telling them to recommend or however it works, wherever somebody is. Think through like, well, what are the effects in the other areas of this client's life, both now and in the future? We have a strong tradition here in the US of a lot of insurance companies leading with insurance products. There's not anything wrong with leading with insurance products inherently.

58:40If we are not thinking about how a whole life policy or universal life policy affects somebody's cash flow now and how they prioritize things, then I think that we're definitely not providing a lot of value. So I think the first thing is to think, okay, well, here's a strategy we would normally use in this particular fact pattern. What are all of the effects that it can have for this particular client? That's one piece of it. Because we're not going to get their values initially. And when people present, they give you your financial information. So you do already have to start thinking about that stuff.

59:13But also along the way is to start thinking about how can we identify what this person values. I had a student read the paper, requested to talk to me. And he mentioned in his email that he thought that if we're bending to somebody's priorities, then they can actually end up with ill-being. So they end up worse off. And the example he used was, well, if somebody is looking for prioritizing financial success, that could theoretically affect their mental health and it can affect all these other things. And so my question to him was, okay, well, is financial success the actual need and value? He got it.

59:53He realized, okay, yeah, well, maybe... Because the financial success is like a strategy. Financial success is a strategy to solve something else. So what are you really trying to do here? To demonstrate how difficult it is to elicit values from people, train on that. There are a lot of programs that help identify certain priorities. Asking them directly sometimes can be helpful. At least that's a starting point. Asking them questions. Have you thought about your legacy? Have you thought about when you might need these investments, right? So like if you're in a whole life policy and you need some money in a few years, well, maybe that's not the best place for it.

1:00:30But conversely, like maybe it's useful for somebody who's late in life, who needs to pay estate tax or whatever. If I'm thinking about real world scenarios, I want to be thinking about how it's going to affect other things. I want to try to model that to the best of my ability. and then I want to make sure that what I'm presenting is aligned with what the client wants. To your point, Ben, I think it would be a great exercise to walk them through how are they feeling about these things and then how does that relate to how they would feel about them in the future or if certain events happen. And then you've captured the three main portions.

1:01:06I don't expect people to compute a structural tensor or an effective tensor. I don't expect that. But thinking about these things, especially for newer planners, older planners, skilled planners, they're already doing some of these things kind of naturally. It's hard to be successful if you don't know like what your client wants and you're misaligned. So I think they're already doing some of those things. But if you're a newer planner, I just really start thinking about these things. If you're an experienced planner, you're not thinking about them, think about them.

1:01:34Ben Felix:So it's really thinking about how if you're making a recommendation that's designed for whatever, estate planning, it's really thinking through, okay, how does that interact with the retirement goal? How does that interact with cashflow? How does it interact with other investments, risk management, tax, all those pieces together and how they affect each other. And then that is informed by what the client says they care about among those different financial planning areas. And you've basically shown with your model that there can be pretty significant value gain or loss by thinking about those things together properly.

1:02:07Yeah.

1:02:08Ben Felix:Very cool. Can you walk us through an example of the sort of pitfalls or costs or whatever, however you want to frame it, when expert domain specific advice is delivered in silos. So if you have like the investment expert and the tax expert working separately as opposed to integrated. The best one is the Delgado scenario, which is she's a business owner. I see this a lot with like high network people. They've got a person who does like whatever, and maybe they don't have somebody to coordinate at all. So they've got a tax person, they got an investment guy, they've got an insurance guy, they've got a state gal, all those things.

1:02:45They say all of these things. And they go to those people and they say, Hey, here's my situation. I've got this business. Here are the things that I want to do. That person will ask really good questions about their domain. They will ask really good questions about, Okay, well, how is your business structured? How many kids do you have? Is anybody work in the business. They're going to ask all of those things. But they often don't think about what the effects are on any of the other domains. So in the Delgado example, the estate planning attorney recommended these family limited partnerships with this nice grass structure.

1:03:24It's going to save like$1.8 million in taxes. But this particular client was prioritizing cash flow. She started this company because she was worried about money. She needed money. And that's kind of never gone away in this particular case. She's really worried about financial security. So moving all this money into these trusts, it eliminated a lot of business distribution. So A for saving taxes, A for estate planning, F for making sure that this client got what they wanted, what they needed. And so that$1.8 million in taxes, that's great. That sounds fantastic. Insurance agent proposed like this huge insurance policy with a big premium, again, cutting into cashflow, which is duplicative of what was going to happen through the estate transfer.

1:04:12Retirement, especially a lot of people say in the US, we always talk about delaying. There's a big debate between delaying Social Security and taking it early. And an economist will tell you like, well, it's dumb to take it early because you get 8 % per year. So you take it early, that's costing you an 8 % with no standard deviation. So other than inflation, actually, even with inflation, because inflation is included typically. So they'll tell you that that's not optimal to take it early. So they're going to recommend taking it to 70. But then in this particular case, the portfolio was depressed because of downturn, which now makes this person have to take withdrawals in a suboptimal situation.

1:04:51If somebody integrated that where there's claiming at 62, retaining some portion of the business for a period of time, insurance that actually fit what was actually going to cover things. Well, now you've got, instead of the 1.8 million estate taxes that are saved or the transfer of risk to this insurance policy that would have saved some money or even transferring some of the risk to later years for investment, the integration premium here was like 2.2 million. And so that's in addition to some of the other savings because it wasn't going to be a full 1.8 million in estate taxes that was say, but some portion was going to be there.

1:05:30So there's 2.2 plus 800 ,000 plus say another 1.2. And so now you're starting to see a lot more value. So that's where the pitfalls of that siloed advice comes from. Firms have goals. If you're an insurance company, your most profitable item is a whole life insurance policy. So I get it. I'm not trying to say you should try to find people who need that. But if you're going to be an advisor and you're going to be somebody who is going to have a fiduciary obligation to a client, you have to think about all of these things. And even people who work at the big brokerages, they don't say they're investment people.

1:06:06They say they're full financial planners, but they lead with investments. Definitely think about that as part of the pie, but we should be thinking about how this affects other things. And I think what ends up happening is it's like, well, we need to get assets in the door. So you need to generate assets. And one way to do that is to say, well, here's how our investments have performed over 10 years. assuming that we can actually continue to do that for the foreseeable future.

1:06:29Cameron Passmore:Does this value of integration and financial planning scale linearly with wealth? I don't know. It's a little bit more nuanced than that. I think it scales with complexity. A very simple thing where somebody is served in the military, they're going to get a pension, they're going to get social security, they don't have any assets, so there's not really a whole lot of estate to plan. That premium is going to be a lot lower. They probably need some insurance. They need to make sure they have enough cash flow. But the couplings are going to be a lot weaker. On the other end, if you do have something, it could be wealth.

1:07:01It could just be complexity. If somebody is a gig worker and they got 20 different gigs, and then those turn into like, well, now I've got four Airbnb properties and I've got some people working for me who are driving. I transitioned my Uber business out of Uber, but I'm still providing driving services. But I've got three people who work for me to do that. that complexity, the value, I think, increases when you have more of that complexity. It's intuitive. It also kind of bore out in the examples. As things were more complex, there was more value generated.

1:07:37Ben Felix:What do you think the theory suggests about the role of human financial planners? Well, that they're necessary. We have a hard time articulating what our goals are, even what we value. We go through life, we take hits, we move, we very seldom have time to think those things through. And I think one of the things that planners do is that they can think about the various consequences for certain actions. And we have training. We have financial training. We have... It's not really legal training, but we understand how a lot of these structures work. We pay attention to the economy. So we understand a lot of the exogenous issues, but then we also get to know people on a personal level.

1:08:21And I think what happens is people in general, they think that some of these big macro things are things that they always need to be concerned about or it's going to affect everything they do and that there's no way to really plan for it. Or that, hey, well, I listened to XYZ Influencer and I've already got it figured out I'm going to be rich sort of thing. What's missing is the fact that a lot of this, I think, is done intuitively. These ideas came from practice. I've practiced for years. I've been doing this almost 21 years. And so I've observed what happens and what I do and how I've changed how I practice to account for a lot of these things.

1:09:03What humans do is we have that ability to talk to a client, look at a situation and realize like, okay, we'll hear the choice that we can narrow the choice set significantly. It becomes much more manageable for people. I think a lot of times people come in and especially, even wealthy people, they come in, they've got a massive choice set. I have a client who his anxiety is so high because his choices are like, at this point in his life are basically unlimited. And like me, you just got to start making decisions. Not that we're like going to fix anxiety or anything like that. But if we can help people narrow their choice set, become more optimal for them, people talk about the optimal outcome.

1:09:47And so a lot of times that's like, it's wealth maximization or higher investment return. And one of the things I tell students is that optimal is not always best. And so what's best for a client may not be what you think or what I think is the most optimal or what an economist thinks or anybody else who does optimization for a living. This essentially describes the complexity of what we do and why it has to be a human. to do these things.

1:10:15Cameron Passmore:So AI is the hot topic these days. From your perspective, how does AI interact with the value of a human advisor? AI is pretty useful in a lot of ways. It can create some content. It can be a good thinking partner to think through certain problems. It's not going to think of all of the consequences for a particular course of action. So there are do-it-yourself people who probably can do this. So I don't want to be like, hey, they can't just use AI to come up with a good financial plan. A, there's a lot of friction there because now you got to check in with the AI all the time about what's happening.

1:10:51The AI has to ask you some of these questions. Maybe you could program something that says, hey, we're going to update your discount rate. We're going to update that today. Maybe that can happen. But also people have lives to live as opposed to the planner who's like, hey, we're going to dedicate this time to deal with these XYZ functions. that's a big part of where I think humans can have that value. But we could use the LLMs to be like, okay, look, I've got this client, here's the situation. And then when it spits something out, you can then think of, here are the second, third, fourth order effects that the model didn't think of or didn't want to expend the token to provide you with that information or both.

1:11:32It can make for a better product. I don't know if it really makes people more productive because my wife and I both have been using it a lot for a lot of things. And sometimes we end up with the same result, but also we spend a lot of time fixing things, asking questions, checking. Not that these errors are massive errors. It's just that if we delegate the whole thing to AI, it's not going to think of the follow-on things to do. It's not going to be monitoring things because it's not a person. It's not a real thinking machine. It's not proactively thinking about what happens. There are times where I'll just sit in my office where I'm thinking about a particular, or up at night thinking about a particular client case, something that I hadn't thought of or some sort of epiphany, or I realized, okay, what they told me about what their priorities were is not exactly accurate.

1:12:24How can I ask them in a way that doesn't get them defensive? All of those things that I do not see AI doing, but I do seeing it say, hey, look, this is the situation as I see it and then it giving me some information that I can be like, oh, okay, yeah, I think that makes sense. Or that's a good direction but let's think about it this other way. This is how I think about AI. This is how I think about how useful it can be. Avengers Endgame, Tony Stark, there's a scene where he figures out time travel, right? He's talking to the AI, he tells it what to do and it outputs basically the model for time travel or whatever.

1:13:04But it was his insights that crafted. Now he had to go back and check because you can actually see after that scene, he's like looking at a few things. He's double checking to make sure that the archers on intelligence actually did it right. And then he tells his wife that he solved time travel. That's how I think about it. I think it can be that useful if we know how to do it. But it's not going to solve that. It's not going to solve something like that on its own. And I think it's the same thing with client. I can ask it some things that are very specific to a client situation and I can probably get some good insights.

1:13:37And then I can say, well, let's do this instead and let's see what happens. One of the interesting things, I think, especially around tax planning, especially around things where there's some ambiguity in the tax code. Here, there's a lot of ambiguities for certain things like business owners. There's a rule, but that rule is not codified in any way that you can quantify what somebody should do. But you can get information about it. We have this thing called an S-Corporation. An S-Corp allows you to avoid payroll tax on distributions. But the IRS is like, Hey, if your salary component is not high enough, then we're going to come get you.

1:14:16But there's no threshold. There's no number on that. They say it has to be reasonable. So reasonable means a lot of things to a lot of different people. So you can actually have the conference and say, Okay, well, these are my job functions, or this is the client's job functions. Get me some information on what that job function costs in this area or whatever. Maybe they're a CEO, they're a financial planner, they're also the custodian. They may be all of those people rolled in one. So their value is split among those things. And so building in, well, what is the salary component versus what is the distribution component now becomes a much more complex problem.

1:14:53The LLM is not going to know. It's just going to say, well, it could be too high, it could be too low. but you can use it. So there are a number of things like that, I think, in planning that make us way more valuable if we can use the tool. So that's my personal take on it. It can be useful. I certainly don't see it replacing it. I just don't understand how that would occur, I guess.

1:15:15Ben Felix:All right, a last question for you, Mike. How do you define success in your life? This paper has been kind of a journey in that. So success for me is being able to, on a whim, stop the day, and go sit on the deck with my wife and have a beverage, provided the weather is not as cold as it is in Canada. Or, you know, by the fireplace, which probably would be more appropriate for you guys. And to be able to spend time thinking of things like this, to spend time in an endeavor that engages my brain, to me, that's success. If I have the time to do those things, then I'm successful. That's a great answer.

1:15:54Ben Felix:Awesome. All right, well, I don't think listeners are going to be disappointed in the level of nerdiness in this episode. So we appreciate that, Mike. Thanks a lot for coming on the podcast to talk about your paper. Thanks. I really appreciate it.

1:16:53Cameron Passmore:Thank you. But that's not the same thing as telling you to sell them. This communication is distributed for informational purposes only. The information contained herein has been derived from sources believed to be truthy, but not necessarily accurate. We really do try, but we can't make any guarantees. Even if nothing we say is fundamentally wrong, it might not be the whole story. Furthermore, nothing herein should be construed as investment, tax, or legal advice. Even though we call the podcast your weekly reality check on sensible investing, and financial decision-making, you shouldn't rely on us when making actual decisions, only hypothetical ones.

1:17:33Cameron Passmore:Different types of investments and investment strategies have varying degrees of risk and are not suitable for all investors. You should consult with a professional advisor to see how the information contained herein may apply to your individual circumstances. It might not apply at all. Honestly, you can probably ignore most of it. All market indices discussed are unmanaged, do not incur management fees, and cannot be invested indirectly, which is a shame because it would be awesome if you could. All investing involves risk of loss, including loss of money, loss of sleep, loss of hair, and loss of reputation.

1:18:07Cameron Passmore:Nothing herein should be construed as a guarantee of any specific outcome or profit. Past performance is not indicative of or a guarantee of future results. If it were, it would be much easier to be a Leafs fan. All statements and opinions presented herein are those of the individual host and or guests, and are current only as of this communication's original publication date. No one should be surprised if they have all since recanted. Neither One Digital nor PWL Capital has any obligation to provide revised statements and or opinions in the event of changed circumstances. See you next time.