I-Orp vs. ********.com

[lawman3966]

when I run I-orp with all the same data (at least, data that is common to the two calculators) as ********, I-orp yields an annual expenditure level about $9K higher than what ******** provides (about 66K instead of 57K (when using the 95% confidence level in ********).

I entered into I-orp what I thought were modest numbers for stock returns, bond returns, and inflation, namely: 6%, 2.5% and, 2%, respectively.

******** uses aftcasting with historical market data. If the I-orp Monte Carlo option is not selected (I did not use this option in I-orp), perhaps I-orp uses some sort of constant return for the stock and bond markets? If true, this would explain the difference, as the 95% confidence level on ******** allows for the possibility of some pretty low stock market returns, presumably lower than 6%. At least, that's my best guess so far.

However, if any of the learned members of this board have a better explanation, I would love to hear it. Unless I learn otherwise, I will stick with the more cautious approach of ********, though I'd love to believe the more optimistic number that emerged from I-orp.

 

[REWahoo]

FYI, ******** is an unauthorized ripoff of FIRECalc and as a result, I do not trust the calculator. Not sure of the creator's motivation, but I am fully aware of his lack of ethical standards.

 

[samclem]

How does your I-orp run compare to FIRECalc?

With FIRECalc you know what the data set is (actual historical US asset class returns). A MC sim tries to replicate some sort of return data, but few of the online calculators actually explain how the soup is made, if the relationships between the asset classes that exist in the historical record were maintained, etc.

 

[orplanner]

How does your I-orp run compare to FIRECalc?

With FIRECalc you know what the data set is (actual historical US asset class returns). A MC sim tries to replicate some sort of return data, but few of the online calculators actually explain how the soup is made, if the relationships between the asset classes that exist in the historical record were maintained, etc.

samclem is seeing is called variance drag. ORP is assuming perfect information in the form of constant, average rate of return and the compounding of asset returns occurs at maximum potential while Firecalc, as a probabilistic simulator, will get lower compounded returns because some random returns are less than the averagerate of return. This is called variance drain in the literature.

A second reason that ORP will come in high is that it is optimizing savings withdrawals to maximizing spending so, given a set of random returns, ORP will return a higher spending value than the simulator. In the end the simulator accepts a spending amount and computes when the money runs out. With or without including income taxes in the model, measuring success or failure by the size of the estate is inefficient as compared to maximizing, optimal spending.

ORP's Monte Carlo method uses historical data. For a full description of how ORP does its Monte Carlo method please see:
The Optimal Retirement Planner (ORP) contains a Monte Carlo linear programming (LP) implementation of a retirement planner

For a critique of Monte Carlo financial techniques please see:
Monte Carlo Retirement Calculators