Modeling a multi-account portfolio?
Posted: Fri Mar 12, 2021 2:31 pm
Recently I've been taking a look at a few different retirement planning applications. One issue I've run into is how to model a portfolio made up of multiple accounts that have wildly varying individual asset mixes.
In the past when I've used one of the many simple calculators on the web to just get a SWAG, I have used one lump sum for the current portfolio value and have input the average expected return for my aggregate asset mix (which I can find from portfoliocharts.com). But more complex retirement planning software allows you to track individual accounts and input unique values for expected return. This allows the software to more accurately model drawdown scenarios, taxes, RMDs, etc.
What I've found is that by using the individual average historical return for each account based on its asset mix, my final portfolio value more than doubles vs taking the average historical return for the entire portfolio and applying it to all accounts. Sounds great at first, but the problem I see with that is that the variance in standard deviation between the individual accounts can be quite high so I'm not sure how much trust to put in the results. For example, the standard Permanent Portfolio has an average return of 5.2% with a standard deviation of 7.1, but an account with only gold has an average return of 6.5% and a standard deviation of 24.7! That's an extreme example but hopefully it illustrates the point.
Just wondering what approach others are taking to modeling different accounts into the future. I believe the saying that "all models are wrong, but some are useful", so I'm not planning to rely too much on a specific result. I'm just trying to build a useful model that I can adjust as time goes on.
In the past when I've used one of the many simple calculators on the web to just get a SWAG, I have used one lump sum for the current portfolio value and have input the average expected return for my aggregate asset mix (which I can find from portfoliocharts.com). But more complex retirement planning software allows you to track individual accounts and input unique values for expected return. This allows the software to more accurately model drawdown scenarios, taxes, RMDs, etc.
What I've found is that by using the individual average historical return for each account based on its asset mix, my final portfolio value more than doubles vs taking the average historical return for the entire portfolio and applying it to all accounts. Sounds great at first, but the problem I see with that is that the variance in standard deviation between the individual accounts can be quite high so I'm not sure how much trust to put in the results. For example, the standard Permanent Portfolio has an average return of 5.2% with a standard deviation of 7.1, but an account with only gold has an average return of 6.5% and a standard deviation of 24.7! That's an extreme example but hopefully it illustrates the point.
Just wondering what approach others are taking to modeling different accounts into the future. I believe the saying that "all models are wrong, but some are useful", so I'm not planning to rely too much on a specific result. I'm just trying to build a useful model that I can adjust as time goes on.