Different results obtained with equivalent inputs

Post questions about how to use the planner and how the planner works as well as comments and suggestions. (If in doubt, post here.)
Post Reply
sidchem233
Posts: 2
Joined: Mon Apr 11, 2016 3:35 pm

Different results obtained with equivalent inputs

Post by sidchem233 » Mon Apr 11, 2016 3:48 pm

I have tried setting the Portfolio return at 5% with a 8% SD versus setting Taxable Return, Tax Deferred Return, Tax Free Return all at 5% with a 8% SD and get a different planned expenses amount (with a activated Goal Seek capability). I get a 7% higher amount for the expenses when I set the individual category returns. Am I making a mistake or is there something in the program that changes when the individual category returns are used.

jimr
Posts: 558
Joined: Thu Feb 28, 2008 6:48 pm

Re: Different results obtained with equivalent inputs

Post by jimr » Mon Apr 11, 2016 5:01 pm

It may be counter-intuitive, but what you're seeing is a function of the way separate returns work. It happens without goal-seek too.

If you create a new plan and enter $300k for the taxable, tax deferred, and tax free portfolios, leave everything set to the default, then go into additional inputs and create 4 portfolio return entries (1 for the overall portfolio, then one each for taxable, deferred, and tax free), you can see the same thing. If you run the plan with only the overall portfolio return entry enabled, you'll get roughly a 65% probability of success. With the three per-portfolio type returns enabled, you'll get a probability of success of around 82%.

The issue is that when you specify separate return parameters for each portfolio, the 3 return sequences are run independently and the volatility of each sequence gets damped out by the others. This gives the overall result much less volatility.

In the above example, you'd need to set the std dev of the independent portfolios to around 12% to offset this effect and get back to a 65% probability of success.

I could easily change the simulation to generate just one random number and use it for all three of the per-portfolio returns. This would 'lock' all the three returns so they move exactly together and only differ in scale. The trouble is that some folks want the three portfolios to move independently rather than in lockstep.

sidchem233
Posts: 2
Joined: Mon Apr 11, 2016 3:35 pm

Re: Different results obtained with equivalent inputs

Post by sidchem233 » Mon Apr 11, 2016 8:38 pm

Jim

Thanks for the reply, I think I understand the difference now and how the feature works with three separate returns. Not sure this applies to the real world very well since there may will probably be correlation between portfolios unless one is very careful to prevent that, not likely in most cases. I may stick to having one portfolio since I generally consider all my investments from different accounts to be in the same portfolio even though I do tend to keep the tax free accounts in higher yielding volatile investments.

Many thanks for the program, I hope it brings you a lot of satisfaction in supporting so many people.

caymann
Posts: 7
Joined: Sun Mar 20, 2016 4:10 pm

Re: Different results obtained with equivalent inputs

Post by caymann » Wed Jun 08, 2016 6:47 am

In the above example a 12% SD is suggested.

Is there a rule of thumb i can use so that the three ( taxable, deferred, and tax free) don't interfere with each other.

In my case, i am looking to do the following...

Taxable portion> age 50-END at (6%/4.3%)
Tax Def portion> age 50-60 at (7%/7.1%) ; age 61-END (6%/4.3%)
Tax Free portion> age 50-60 at (8%/9.9%) ; age 61-70 (7%/7.1%); age 71-END (6%/4.3%)

thanks
jim

jimr
Posts: 558
Joined: Thu Feb 28, 2008 6:48 pm

Re: Different results obtained with equivalent inputs

Post by jimr » Wed Jun 08, 2016 7:12 am

I got to 12% just by trial and error. I don't think there's any rule of thumb per se.

In your case, I'm not really sure if breaking portfolio returns down by portfolio type is adding more signal than noise. I put this feature in after multiple user requests for it, but I have mixed feelings about it because of the subtleties in how it works. In your case, it may be fine to just estimate one blended return/std dev for the three combined portfolios and use that. You can still vary the return/std dev over time as you've been doing with separate returns.

IMO, with this type of analysis there's a risk of the noise getting above the signal if you try to include more granularity or resolution than the inputs support. Most experts feel that this type of long range planning tool has a built in error signal of 10 or even 20 percent, based on the reality that so many inputs are unknowable.

So the key is to consider how best to use the tool for decision support. That can be accomplished by trying to understand the contours of the plan and which inputs affect the plan the most. For example, exploring the results using a std deviation of 10% vs 20%, with all else equal, can tell you how vulnerable the plan is to large market volatility. Another interesting test can be to compare results with greater or lower withdrawal rates to see how sensitive success is to that input.

Hope that makes sense,

Jim

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest