reduced expected return

Post questions about how to use the planner, user inputs, how the planner works, and comments and suggestions.
Post Reply
meeotch
Posts: 2
Joined: Wed Oct 28, 2020 1:25 pm

reduced expected return

Post by meeotch »

So you hear a lot recently about historical average market returns being an unrealistic estimate for the coming decade(s). Low interest rates, "secular stagnation", current historically high P/E ratios, etc. With the caveat that nobody actually knows, I personally believe it to be true, and in my own retirement spreadsheets, I've been using a pretty conservative 5% as an expected return for a 60/40 (plus some cash) portfolio. I've got two questions relative to FRP:
  1. If I set the investing style to Custom, and the avg return to 5%, what should I set the std dev to? A 60/40 portfolio looks most similar to the Moderate Risk preset... So is 9.9% then the appropriate std dev? I.e., fluctuations in the market are just as pronounced as always, it's just the average return that's depressed.
  2. Is the possibility of a "lost decade(s)" (or depressed, anyway) already baked in to FRP by virtue of the Monte Carlo simulation - implying that the 5% I use for my deterministic spreadsheet calculations isn't appropriate for FRP, because FRP is already simulating the chances of starting off with a low-return stretch? (Or maybe I should split the difference, and use something like 6.5%.)
jimr
Posts: 821
Joined: Thu Feb 28, 2008 6:48 pm

Re: reduced expected return

Post by jimr »

I'd say you're asking the exact right questions and focusing on a really important area that's a concern for all of us.

Unfortunately, the answer to your question falls more on the side of financial planning advice than the technical support and how to use the planner side and since I'm not a practicing financial advisor, I'd rather not step out of my lane to give you a more concrete answer.

One online resource that I've found helpful personally is the bogleheads.org forum. IMO, this community has a great track record of providing solid resources and support for DIY financial planners.

One of the users at bogleheads developed a spreadsheet to track historical portfolio returns and this spreadhseet might be helpful to you for trying to determine what standard deviation would be most appropriate to use with your estimated return.

https://www.bogleheads.org/wiki/Simba%2 ... preadsheet

Best,

Jim
meeotch
Posts: 2
Joined: Wed Oct 28, 2020 1:25 pm

Re: reduced expected return

Post by meeotch »

Thanks for the reply. Yes, I did take a look at the bogleheads spreadsheet - if memory serves, the 60/40 allocation gave numbers similar to the Moderate setting in FRP: 9% returns, 10% std dev. Which makes sense, I guess, since the sheet is based on historical data. (Side note: bogleheads is terrific, but every time I post a question I get at least one, often multiple cranky bastards telling me my question is stupid or somehow invalid. Thanks, internet!)

But back to my original query: Am I right in suspecting that the Monte Carlo scheme does "bake in" the probability of bad years - *but*, if one is convinced that returns in the next decade or two are likely to be low, then inputting a lower avg return is appropriate anyway? Reason being that the bad/good years will be fluctuating around an average which is itself lowered.
jimr
Posts: 821
Joined: Thu Feb 28, 2008 6:48 pm

Re: reduced expected return

Post by jimr »

Am I right in suspecting that the Monte Carlo scheme does "bake in" the probability of bad years - *but*, if one is convinced that returns in the next decade or two are likely to be low, then inputting a lower avg return is appropriate anyway?
Monte Carlo bakes in the probability of bad years to the extent that the future looks roughly the same as the past. However, if you expect returns over the next decade to be lower than the historic average then the way to model that in FRP would be to reduce the historical average return and/or increase the associated historical standard deviation.

This is an area where sensitivity analysis might be helpful. There's an option in sensitivity analysis to have one sensitivity parameter be return and the other one be standard deviation (I think it's the last item in the drop down). This should give you a visual display of how your plan fares under various combinations of return and std deviation. Just be aware that any rates you set in additional inputs will override the sensitivity parameters (eg if you set up any portfolio return entries in additional inputs, they'll override the variations in the sensitivity parameters).
target2019
Posts: 20
Joined: Fri Dec 20, 2019 8:41 am

Re: reduced expected return

Post by target2019 »

jimr wrote: Wed Oct 28, 2020 5:55 pm This is an area where sensitivity analysis might be helpful. There's an option in sensitivity analysis to have one sensitivity parameter be return and the other one be standard deviation (I think it's the last item in the drop down). This should give you a visual display of how your plan fares under various combinations of return and std deviation. Just be aware that any rates you set in additional inputs will override the sensitivity parameters (eg if you set up any portfolio return entries in additional inputs, they'll override the variations in the sensitivity parameters).
I've been using 5% Return - Average and 10% Return - Std Dev since looking in the Simba spreadsheet (available from Bogleheads).

Based on your guidance above I used Sensitivity Analysis to explore Return - Avg vs. Return - Std Dev. It's an eye opener as to how that setting affects Probability of Success.

Now I wonder if I should go further and increase 10% Return - Std Dev to something like 15%? Of course that gets into financial planning.

But I'm also wondering about a programmatic thing, namely, does Return - Avg go negative in some runs when the Std Dev (say 10% setting) exceeds the return setting (5%)?
jimr
Posts: 821
Joined: Thu Feb 28, 2008 6:48 pm

Re: reduced expected return

Post by jimr »

The short answer is that yes, as long as the standard deviation is high enough, the randomly selected return will almost certainly be negative in some years of some simulation paths. In the example you cited, I'd guess the simulation would draw negative returns thousands of times during a single run.

Here's a Wikipedia page that explains how this works:
https://en.wikipedia.org/wiki/68%E2%80% ... 399.7_rule

It turns out, the standard deviation doesn't even need to be as high as in your example for there to be a good chance of negative return being generated in at least some years of some simulation paths.

As that Wikipedia article explains. There's a roughly 5% chance that a randomly selected return will fall outside of a 2 standard deviation range around the average. That means there's a roughly 2.5% chance of drawing a return that's lower than the average return minus twice the standard deviation. (note that half of the 5% of returns outside the range will be less than the average return and the other half will be greater than the average return)

This means with a return of 5% and a standard deviation as low as 2.5%, you'd likely still draw thousands of negative returns in each run. Say your plan has 30 years in it and you do a standard run with 10,000 simulation paths. The simulation will draw 300,000 random returns during the run. With a 2.5% chance of drawing a return lower than the average return minus two times std dev (eg lower than 5% - 2 x 2.5%), you'd likely draw negative returns 7,500 times (300k x 2.5%).

Hopefully I did the math right on that example :)
Post Reply

Who is online

Users browsing this forum: No registered users and 24 guests