Six Sigma Analysis and Ibonds

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
tindel
Posts: 149
Joined: Sun Nov 12, 2017 10:06 am

Six Sigma Analysis and Ibonds

Post by tindel » Wed Mar 13, 2019 2:02 am

Today (well, yesterday now) the CPI-U index was released for February - which isn't really big news, but it gives a new data point into what the inflation portion of the I-bond rate will be in the next cycle. Since I'm studying for my Lean Six Sigma Green Belt cert, I wanted to attempt to merge my newly acquired knowledge with something more practical and possibly useful. I decided to analyze what the chances are that the next I-bond cycle inflation adjusted interest rate would be 0.00% or less. It's worth noting that the inflation adjusted rate can be negative, but your composite rate can't be below 0.00%.

More interesting than the release of the new data is that there was a period of deflation over the holidays, so this is a particularly interesting 6 month period where we have a greater likelihood of seeing a period of deflation for the next 6 months than in recent past. I did the math and found that March will need to deflate the index by -0.13% to provide us a deflationary period. (The CPI-U will need to be 252.451 or less).

I looked at the percent change each month from the period of Jan '16 to Feb '19 - a little over 3 years of data points. I determined that the monthly percent change is likely a normal distribution using a chi-squared goodness-of-fit analysis. With this knowledge I was able to use the normal distribution to determine the probability that the next I-bond cycle will be deflationary.

The result? There is a 9.38% chance that we will have a 0.00% or lower return for the inflation adjusted portion of the I-bond interest rate.

Interesting but not yet actionable:
A period of deflation would give some interesting advantages to those that have low fixed rates and held less than 5 years. They may have an opportunity to cash out without taking the 3 month interest penalty in order to attempt to acquire a higher fixed rate in the future.

Actionable item:
At this point some people may want to consider holding off on buying their allotments until the next 6 month cycle rolls around in November due to probable rates being lower than savings account yields. Of course, you could lose out on the relatively nice 0.5% fixed rate available now.

More interesting stats:
The chance that the semi-annual rate stays at the current rate of 1.16% or better is only 0.013%
The chance that the semi-annual rate is lower than 0.98% (the lowest in the last 3 years - March '17) is 99.794%
The 75% confidence semi-annual rate for the next 6 months will be 0.31% +/- 0.27%.
The +/-3 sigma confidence (99.7%) semi-annual rate for the next 6 months will be 0.31% +/- 0.69%.

User avatar
Svensk Anga
Posts: 587
Joined: Sun Dec 23, 2012 5:16 pm

Re: Six Sigma Analysis and Ibonds

Post by Svensk Anga » Wed Mar 13, 2019 10:04 am

I think there is much less chance of <0% inflation in the next I bond adjustment than you estimate.

What you are missing is that there is a fairly strong seasonal component to inflation. That is why the government always reports seasonally-adjusted CPI. Typically we have deflation in October, November, and December, then the worst inflation of the year January-April. The other five months are mid-range. I think that the treasury deliberately set the I-bond rate-set schedule as they did so as to even out the inflation rate between the two six-month periods. The way they do it, they average the three deflation months with three of the highest inflation months (October - March) to calculate the inflation adjustment effective May 1. The six moderate inflation months typically give a similar rate.

The seasonal effect ought to make the March number greater than zero. Watching gasoline prices can give an early clue. Still, the next six-month rate is going to look pretty poor compared to short-term rates available elsewhere. The composite rate will still be good for the old, high real rate bonds and just okay for the newest 0.5% real bonds. The last few years of 0 to 0.2% real rate I bonds are going to look poor compared to a Vanguard money market fund, at least for six months.

Since 2002, the monthly average inflation rate has been:
0.35% January
0.42%
0.56%
0.38%
0.29%
0.22%
0.05%
0.14%
0.19%
-0.05%
-0.26%
-0.23% December

March has been the worst month in this period

User avatar
jeffyscott
Posts: 8238
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Re: Six Sigma Analysis and Ibonds

Post by jeffyscott » Wed Mar 13, 2019 2:21 pm

The composite rate will still be good for the old, high real rate bonds and just okay for the newest 0.5% real bonds. The last few years of 0 to 0.2% real rate I bonds are going to look poor compared to a Vanguard money market fund, at least for six months.
[/quote]

I think even the 0.5% ones are going to look very poor compared to alternatives, unless cash rates drop a lot. The new variable rate would be 0.26%, if there would be no change in March. Not sure how much we can expect March to add, but guessing the variable is likely to be less than 0.5%, if so then the composite will be less than 1% even for the 0.5% I-bonds.

I've been on the fence about selling our 0%-ers for a while. Now with what is likely coming up in May, I'll be getting rid of not only those, but all except the few oldies that are at 3%+.
Time is your friend; impulse is your enemy. - John C. Bogle

statman
Posts: 98
Joined: Mon Jul 20, 2009 6:07 pm

Re: Six Sigma Analysis and Ibonds

Post by statman » Wed Mar 13, 2019 3:32 pm

My goodness. The chi-square test (Karl Pearson, 1900) is not at all suitable for assessing normality. To start with, it breaks the continuous distribution into discrete cells. Get some semi-modern software and try Kolmogorov-Smirnov or Anderson-Darling. And don't expect to significantly reject normality with 36 data points. And monthly CPI values are most likely serially correlated rather than than independent, which is confounded with lack of normality for all common tests of fit. OK, stat prof blows off steam.

dbr
Posts: 30561
Joined: Sun Mar 04, 2007 9:50 am

Re: Six Sigma Analysis and Ibonds

Post by dbr » Wed Mar 13, 2019 5:10 pm

Maybe it would be good to read here: https://www.amazon.com/Misbehavior-Mark ... oks&sr=1-2

That is not to say that Mandelbrot actually arrived at solutions to any of the issues he points out.

Post Reply