dory36: old vs. new firecalc results?

[jdw_fire]

The range of annual numbers discussed have been in the 3-5% range for inflation and in the 8-12% range for growth.

Having read a number of posts on this board I suggest that the growth rate for money to cover current year expenses would be more on par with the rates produced in a MMA and thus much closer to the inflation rate.  If this is so then the growth is cancled out by the increas in inflation through the year.

 

[modlair]

Thanks for giving it thought and effort, dory. I'm gonna scratch my head some and think. I was thinking originally that a relatively small number in year 0 would compound over 30 or 40 yrs and get pretty serious later, but you may have shot that down.

Hmmm.

 

[dory36]

A small amount can grow a whole lot, but this isn't a linear function. All we care about is how many cycles dip below zero.

 

[modlair]

Okie doke, Sunday morning and time to think. Let me get the core concepts understood:

1) We're looking for an inaccurate number of failed cycles. If we adjust a total on the portfolio with some change in the calculation, but that adjusted total is not sufficient to change the number of failed cycles, then it is nitpicking even if the portfolio value changes what appears to be a significant amount. Nothing is significant and wastes your time unless it changes the number of failed cycles. I think this is clearly true. The number of failed cycles is what determines confidence level so that is all that matters.

2) Apparently your implementation of inflation on spending is done in January, but uses December's inflation measure (??). This means that for year 1876, you apply the (Dec) end of year inflation measurement to the year's spending -- even though you're going to pre-spend that money in January. And because you prespend that money in January, that year's spending does not reside for any months in an interest bearing account and does not grow.

3) I said 1876 in item 2) but you also do the same thing in year 0 (1871?). 1871's end-of-year reported inflation determines the spending level of 1871, after being added to the user's input number for first year spending. So if a user says his first year's spending will be 30,000, you will actually spend in that first year (30000 X (1.0 + infl rate)).


Okay jdw_fire pointed out that if we presume that short term rates are pretty close historically to inflation, anything done to say that spending is spread over the year means that inflation cancels growth. I think this is true. The only question is does "inflation cancels growth" constitute an increase or decrease in portfolio size in comparison to the current algorithm (and of course if that size change is big enough to generate new cycle survivals/failures).

dory, your post seemed to try to explore any difference in inflation vs short term returns as the year traverses and yeah, that's pretty convincing that such an amount will be small (though it will compound as an annuity, with year 0 compounded 30 yrs, year 1 compounded 29 yrs etc), but I suspect it is not going to generate a cycle survival/failure change so that is persuasive.

The only remaining doubt I would suggest is that a user specifies a first year spending value and that is not what he or she gets. He gets the number he or she specified plus that year's inflation. This difference is not going to compound (I don't think), but it means one year's worth of inflation will affect the successful SWR. I'm guessing small potatos and at most this might change one cycle, but I'm not sure.

Anyway, thanks for entertaining the questions.

 

[dory36]

A few points, for those lost in this thread.


1 - The user enters what it costs them today to live on, annualized, not what they think it will cost them to live on in the coming year. They don't and can't know that. Because FIRECalc uses historical data, it can use the year end results, even though they are not knowable by the retiree in January.

2 - As a failure-avoidance utility, FIRECalc and other similar tools don't particularly care if a strategy will make you enough money to hire Bill Gates as a houseboy -- that is still simply a "1" in the number of successes column, the same as if you had ended with 12 cents. The only number that matters in the failure avoidance tool is the worst case outcome. If the worst case is greater than zero, it's a success.

3 - FIRECalc and similar tools are planning tools, not spending models. After using FIRECalc and other tools to decide that living off your portfolio and other non-employment sources is feasible, you don't ignore the world. Your actual spending model is likely to be more like ESRBob's 95% rule, where you party hearty following great years, and tighten your belt when the market, unexpected expenses, or anything else makes you think that doing so is a good idea.

 

[dory36]

A Memorial Day present for rodmail...

I modified the code so the inflation-adjusted withdrawal for year 1 is the average of the nominal withdrawal and the year-end inflation-adjusted amount.

The difference is quite a bit less than I expected. In the default conditions, the success rate is unchanged, and the average ending portfolio increases by only $992, after 30 years. The withdrawal amount for a 100% safe W/D goes up only $19 a year.

But it is calculated more closely to actual likely spending patterns, spread throughout the year, so the idea makes sense to incorporate.

OTOH, factoring in growth of money taken out of the portfolio is not something I feel comfortable doing. I don't know whether or how people would invest these withdrawn funds, and I have no good source of data for the sorts of very short term investments that the average investor would have had access to in 1871 and later.

So, once money is withdrawn, it is "out of sight" to FIRECalc.

 

[modlair]

A Memorial Day present for rodmail...

haha, excellent! I'm a former USAF officer so my fellow servicemen thank you.

Not surprised it proves insignificant. The discussion was headed in that direction.

Now . . . about tax loss carry forwards . . . .