4 ms·

I've long been interested in an investment company that had a variable cut that was directly based on the customers' preferences. Ultimately you'd want the mana

by ikeboy 10y ago

I've long been interested in an investment company that had a variable cut that was directly based on the customers' preferences. Ultimately you'd want the managers incentives based on the fees to be exactly aligned with the customers' needs, which means that different fee structures are needed for different kinds of customers. Basically the hedge fund fee structure extended. So someone who wanted more risk would give a higher cut of fees for higher gains and a lower cut of fees for lower gains, and vice versa. The average expected value to the manager should be a constant percentage of assets, but the distribution changes for each customer. One cool thing that naturally falls out of this idea is negative fees: if someone is risk averse enough, then the incentives require the manager to lose money if the customer loses, which causes the manager to be risk averse as well for those funds. (I have more detailed thoughts on this that this margin is too small to contain; feel free to email me for some disorganized elaboration on the above.)

kaistinchcombe 10y ago
The place this ends up sticky is that a manager with multiple clients can jointly maximize fees – e.g. instead of properly hedging a bet within each client's portfolio, you can put the bet in one client's portfolio and the hedge in another's. If risk preferences are divergent enough the manager can basically be making one-way bets – and frankly the point of this is it could be socially efficient for two of your clients to bet against each other, that's the point, but for almost any nonlinear compensation structure you dream up (even just sharing a little bit in upside, like venture) you can figure out that the manager maximizes his/her own risk adjusted return by playing clients off against each other or creating scenarios that sometimes throw a client under the bus. To be clear, a flat fee is the ULTIMATE one-way bet… but making the one-way bet simple rather than complex means that it doesn't incentivize one type of behavior over another, you get the one-way bet even without throwing a client under the bus. And maybe at the end of the day your broker or advisor is a good human and, barring any incentive to the contrary, will do his/her best for the client. Kind of a crazy bet, but…
- ikeboy 10y ago
  >frankly the point of this is it could be socially efficient for two of your clients to bet against each other So I did think of your scenario, and actually one worse (directly moving money from clients who don't value the next marginal dollar as much as clients who do). A couple of points I thought of in response: 1. Require clients funds to be transparent, so no moving money around after the allocation decisions have been made (this protects more against my problem than yours) 2. Assume the manager's utility is linear in money, in which case hedging their own fees doesn't make sense. If it's not linear in money, then this all becomes more complicated. I got to the point of saying something like "get a model for client utility and for manager utility of various amounts of money, then set the fee to align incentives". If the manager's utility is linear, then fees should be proportional to utility the client gained. If not, it becomes more complicated and perhaps gaming like you describe is inevitable, this is ultimately a math question for which I haven't modelled everything out and so don't know. You may be able to still define a fee structure that gives a proportional increase of utility to the manager for an increase of utility to the client, but I don't know for sure how that would work for multiple clients. 3. If fees represent utility, and utility is maximized, then wouldn't it be good anyway? (Kind of the point you made in the part I quoted.) 4. Another point to keep in mind is how various kinds of returns are expected to attract or repel customers, which factors into manager utility in a sort of "side-channel". If that's significant enough, it might make the whole incentives not work. There are a lot of interesting math questions that go into this. I don't have too many answers, but I feel like I have a good grasp of the questions. Some other points I see in my notes on this: 1. it could be that various kinds of existing funds already cover most of what investors really want, and this won't be enough of an improvement to justify the overhead. Investors who want a small chance of a large win go to hedge funds, investors who want the whole market buy index funds, etc, could be there's no room for innovation there 2. incentive structures need to be very specified, won't just be a fee for various amounts but need to take time dimension into account 3. For that matter, utility is time-sensitive, so we need a complex model just to determine utility, and maybe quizzing investors on utility won't actually get the "truth" I like to think of this as a combination of the insights behind insurance and hedge funds, basically skewed utility curves. I'd love to talk further if you're interested. I ended up starting a retail startup instead of working on this idea because it's easier to break into, but I'd be thrilled if someone turned it into a real company or incorporated it into an existing one, and you guys seem to be capable of doing something like that.

by ikeboy 10y ago

kaistinchcombe 10y ago

The place this ends up sticky is that a manager with multiple clients can jointly maximize fees – e.g. instead of properly hedging a bet within each client's portfolio, you can put the bet in one client's portfolio and the hedge in another's. If risk preferences are divergent enough the manager can basically be making one-way bets – and frankly the point of this is it could be socially efficient for two of your clients to bet against each other, that's the point, but for almost any nonlinear compensation structure you dream up (even just sharing a little bit in upside, like venture) you can figure out that the manager maximizes his/her own risk adjusted return by playing clients off against each other or creating scenarios that sometimes throw a client under the bus. To be clear, a flat fee is the ULTIMATE one-way bet… but making the one-way bet simple rather than complex means that it doesn't incentivize one type of behavior over another, you get the one-way bet even without throwing a client under the bus. And maybe at the end of the day your broker or advisor is a good human and, barring any incentive to the contrary, will do his/her best for the client. Kind of a crazy bet, but…

ikeboy 10y ago
>frankly the point of this is it could be socially efficient for two of your clients to bet against each other So I did think of your scenario, and actually one worse (directly moving money from clients who don't value the next marginal dollar as much as clients who do). A couple of points I thought of in response: 1. Require clients funds to be transparent, so no moving money around after the allocation decisions have been made (this protects more against my problem than yours) 2. Assume the manager's utility is linear in money, in which case hedging their own fees doesn't make sense. If it's not linear in money, then this all becomes more complicated. I got to the point of saying something like "get a model for client utility and for manager utility of various amounts of money, then set the fee to align incentives". If the manager's utility is linear, then fees should be proportional to utility the client gained. If not, it becomes more complicated and perhaps gaming like you describe is inevitable, this is ultimately a math question for which I haven't modelled everything out and so don't know. You may be able to still define a fee structure that gives a proportional increase of utility to the manager for an increase of utility to the client, but I don't know for sure how that would work for multiple clients. 3. If fees represent utility, and utility is maximized, then wouldn't it be good anyway? (Kind of the point you made in the part I quoted.) 4. Another point to keep in mind is how various kinds of returns are expected to attract or repel customers, which factors into manager utility in a sort of "side-channel". If that's significant enough, it might make the whole incentives not work. There are a lot of interesting math questions that go into this. I don't have too many answers, but I feel like I have a good grasp of the questions. Some other points I see in my notes on this: 1. it could be that various kinds of existing funds already cover most of what investors really want, and this won't be enough of an improvement to justify the overhead. Investors who want a small chance of a large win go to hedge funds, investors who want the whole market buy index funds, etc, could be there's no room for innovation there 2. incentive structures need to be very specified, won't just be a fee for various amounts but need to take time dimension into account 3. For that matter, utility is time-sensitive, so we need a complex model just to determine utility, and maybe quizzing investors on utility won't actually get the "truth" I like to think of this as a combination of the insights behind insurance and hedge funds, basically skewed utility curves. I'd love to talk further if you're interested. I ended up starting a retail startup instead of working on this idea because it's easier to break into, but I'd be thrilled if someone turned it into a real company or incorporated it into an existing one, and you guys seem to be capable of doing something like that.