When I run the historical analysis, my ending portfolio balance is roughly what I've seen with other calculators (ie. firecalc).
When I run the monte carlo option, the ending balance is roughly half of the historical value.
What might I be doing wrong?
Can you please give more details about your monte carlo parameters? (Withdrawal method? Simple or advanced portfolio modeling? Asset allocation? Expected returns?)
Can you please give more details about your monte carlo parameters? (Withdrawal method? Simple or advanced portfolio modeling? Asset allocation? Expected returns?)
Aren't all those parameters used in the same manner for both historical and MC?
@feh Hi, when running the Historical analysis, historical inflation rates and asset rates of return are used (based on your mapping of your accounts/asset classes to the historical asset classes).
When running the Monte Carlo analysis your own rates of return (randomized) and inflation rates are used.
I suspect your rates of return are materially lower (more conservative) than the actual historical averages. You can confirm this by looking at the Historical Analysis results chart. Does the 'Deterministic Savings' line run below the blue bands? If so, that suggests your own RORs are low compared to the historical averages. Not necessarily a bad thing....
@feh Hi, when running the Historical analysis, historical inflation rates and asset rates of return are used (based on your mapping of your accounts/asset classes to the historical asset classes).
When running the Monte Carlo analysis your own rates of return (randomized) and inflation rates are used.
I suspect your rates of return are materially lower (more conservative) than the actual historical averages. You can confirm this by looking at the Historical Analysis results chart. Does the 'Deterministic Savings' line run below the blue bands? If so, that suggests your own RORs are low compared to the historical averages. Not necessarily a bad thing....
On the Analyze->MC Analysis page, there is a "Rate of Return Std Deviations" tab. I have all asset classes set to 0, as the information note says any value above 0 will override the default.
If that note is accurate, I should be using the default rate of return std deviations, whatever they are.
@feh The "Rate of Return Std Deviations" tab allows you to enter custom rates of return. These are not used in the Monte Carlo unless you check the "Use Custom ROR Std Deviations" box at the top of the Monte Carlo Analysis page. This box is not currently checked for any of your scenarios.
But if you did check it and re-run the MC, you would find that the pop-up tooltip's message "any value above 0 will override the default" is wrong. 0 values (and blanks) are indeed accepted as your override Std Deviation values and used in the Monte Carlo Analysis which I just tested and confirmed. When the ROR Std deviations are 0, there is no randomization applied to your account or asset class RORs and so each of the 1,000 Monte Carlo iterations produces the same result as your 'deterministic' projection. All the M/C percentile bands collapse into a single line behind the deterministic line.
I will change the pop-up tooltip message in the next release. Also I will add a better explanation on the "Rate of Return Std Deviations" tab than just "See the manual for details."
But, let's return to your original question:
When I run the monte carlo option, the ending balance is roughly half of the historical value. What might I be doing wrong?
In the little table above the Monte Carlo and Historical Analysis charts there is a column showing the median final savings for the 1,000 Monte Carlo iterations and the 75 or so Historical Analysis projections. As you said, the M/C median is about half of the Historical median. The explanation is: the rates of return you defined on your asset classes is, in aggregate, lower than the historical rates of return.
On the Monte Carlo chart, the deterministic solid line runs along the middle of the percentile bands.
On the Historical chart, the deterministic solid line runs way below the percentile bands, confirming that your portfolio rates of return are lower than the historical averages.
The other explanation I can think of is that your asset class mapping to the historical asset classes is off, but I checked it and it looks reasonable.
I should be using the default rate of return std deviations, whatever they are
The default standard deviation is 1.5 * the asset class's nominal rate of return. There is currently no page in Pralana that shows these values, so I will add it to the Review > Tab Projections > Portfolio tab (or similar place) some time soon.
@feh The "Rate of Return Std Deviations" tab allows you to enter custom rates of return. These are not used in the Monte Carlo unless you check the "Use Custom ROR Std Deviations" box at the top of the Monte Carlo Analysis page. This box is not currently checked for any of your scenarios.
But if you did check it and re-run the MC, you would find that the pop-up tooltip's message "any value above 0 will override the default" is wrong. 0 values (and blanks) are indeed accepted as your override Std Deviation values and used in the Monte Carlo Analysis which I just tested and confirmed. When the ROR Std deviations are 0, there is no randomization applied to your account or asset class RORs and so each of the 1,000 Monte Carlo iterations produces the same result as your 'deterministic' projection. All the M/C percentile bands collapse into a single line behind the deterministic line.
I will change the pop-up tooltip message in the next release. Also I will add a better explanation on the "Rate of Return Std Deviations" tab than just "See the manual for details."
But, let's return to your original question:
When I run the monte carlo option, the ending balance is roughly half of the historical value. What might I be doing wrong?
In the little table above the Monte Carlo and Historical Analysis charts there is a column showing the median final savings for the 1,000 Monte Carlo iterations and the 75 or so Historical Analysis projections. As you said, the M/C median is about half of the Historical median. The explanation is: the rates of return you defined on your asset classes is, in aggregate, lower than the historical rates of return.
On the Monte Carlo chart, the deterministic solid line runs along the middle of the percentile bands.
On the Historical chart, the deterministic solid line runs way below the percentile bands, confirming that your portfolio rates of return are lower than the historical averages.
The other explanation I can think of is that your asset class mapping to the historical asset classes is off, but I checked it and it looks reasonable.
I should be using the default rate of return std deviations, whatever they are
The default standard deviation is 1.5 * the asset class's nominal rate of return. There is currently no page in Pralana that shows these values, so I will add it to the Review > Tab Projections > Portfolio tab (or similar place) some time soon.
Thanks Charlie.
So, to summarize - the MC results are lower simply because the rates of return I specified for asset classes are lower than historical.
ETA - changed rates of returns to historical values and confirmed this was the issue.
