Bootstraping: Analytical Tool

[Gotadimple]

The latest from FundAdvice.com (Merriman-Best) about another analytical technique to determine the probability of having a well funded retirement during the drawdown phase. They use the returns from (I assume the S&P 500) over 39 years to estimate the likelihood of a portfolio lasting anywhere up to 39 years.

It is interesting to read at any rate.

FundAdvice.com - Bootstrapping: Another way to assess potential investment outcomes

-- Rita

 

[OAG]

Wonder how this piece would look if they factored in retiring today after the impact of the past 15 months. I doubt if the hypothetical person in this piece would be feeling safe pulling $137K out of the "revised" portfolio now as the percentage would be about double (close to 10% the first year). Also most of the data that this is based on is pretty OLD. I am older than 65 and do not consider myself in the "distribution phase" as I still "save". However, the piece is interesting for the suggested theory.

 

[2B]

I am older than 65 and do not consider myself in the "distribution phase" as I still "save". However, the piece is interesting for the suggested theory.

That's an interesting observation. The general planning tools somehow look at your retirement date as when you begin the distribution phase. However, thrifty people in retirement still want to increase their nest egg. Old habits are hard to change. After decades of focusing on an increasing net worth, it's hard to accept the slow, continuing drop as a "plan."

 

[LOL!]

I liked this article which came up with a couple of tidbits:

1. Don't have an asset allocation below about 30% in equities or you will run out of money in retirement; and

2. Don't retire in 2008 (or any other bad year for the stock market) or you are screwed. In other words, OAG: they did factor in retiring today.

The 5% withdrawal rate used for the numbers is too high, but they promise future articles looking at different withdrawal rates.

 

[kumquat]

I didn't read it carefully but just skimmed it. Aren't they suggesting they use something similar to FireCalc but with only 39 years of data?

 

[REWahoo]

Aren't they suggesting they use something similar to FireCalc but with only 39 years of data?

Looks that way to me:

Starting with a portfolio of $1 million, the tables show the impact on the portfolio for a variety of asset allocations, using different withdrawal rates adjusted for constant inflation, for the returns which actually took place between 1970 and 2008.

Bootstrapping, as we use it, starts with the actual returns for all the years between 1970 and 2008. We then generate a number of return scenarios based on the number of years of returns we have, in this case 39. The first potential sequence of returns is the actual historical returns, starting with 1970 and going through 2008. Another possible set of returns starts with 1971, goes through 2008, and then loops back again to 1970. A third possible set of returns starts with 1972, continues through 2008, and then loops back to 1970 and 1971 at the end of the sequence, to get the 39 years.

Maybe instead of calling it "Bootstrapping" it should be called "FIRECalc Lite".

 

[jdw_fire]

same as FIRECalc except that they use the same set of years and get variation by starting with a different yr and wrapping around (which is not realistic at the wrap point)

 

[LOL!]

I think the method is about as realistic as FIRECalc and helps to answer questions about retiring at different points in the business cycle. As for the discontinuity as the various wrap-around points (remember one could use different "years of data" against "different wrap points", e.g. 39 years of data with a 20 year wrap around) I did not find that objectionable. The point of the author was that it was an alternative method to be used in conjunction with a Monte Carlo simulation and could be more "random" than the FIRECalc method of just using a single historical sequence of events. I would not be surprised if the method was copied from another field of study that he did not reference.

As for using a single sequence, I recall a comment by a poster here that rebalancing was found in one study to be best every 4(?) years. But what happened is that the data was historical and when that number was used, the rebalancing was nearly perfect around the beginning of the Great Depression or something like that. I would guess that averaging a set of "boot-strapped" runs would have removed this effect of starting at the perfect point.