Although this is not too big a deal, I just want to comment on some arcane side effects of this hole in the data.
Does it Matter for Cost-of-Living Adjustments?
The absence of the data would cause problem for any contracts that use the CPI to adjust pricing. That said, I find it unlikely that there are many contracts that adjust pricing on a monthly basis. It is up to the lawyers to find what the fall back calculation method is.
Under the assumption that data publication resumes for the November data (which is released with a lag), this is only a one-month hole to fill. It would really only be an issue for annual adjustments that are done based on October data.
Does it Matter for Economists?
Adding a one-month data hole to series does not matter that much for business cycle analysis. There is no particular reason that inflation was that interesting in October. If you ran into missing data when you had some traumatic pricing event (like a currency peg snapping) you might have analytical concerns, but the developed economies have considerable inflation persistence in “normal” environments.
For people who deal with economic data, the data hole is a technical problem. Most statistical packages handle missing data seamlessly courtesy of the power of the NaN (Not a Number) value in IEEE floating point data conventions. NaN’s propagate through calculations in a reasonable fashion. The only challenge are things like frequency conversions, such as converting to quarterly — do you remove the entire quarter?
Since the usual convention is to look at 12-month percentage changes, the hole matters for two month’s inflation readings: October 2025, and October 2026.
How to Fill the Hole?
If you wanted to fill in the missing CPI data point because you don’t want to deal with undefined values, one needs to be somewhat careful. Since there is a seasonal pattern to the data, the only simple solution is to interpolate the September and November seasonally adjusted (SA) series. If you want to fill in a non-seasonally adjusted (NSA) point, you then need to reverse the seasonal adjusted pattern to get an implied NSA value. (You can estimate this reverse pattern by looking at the ratio between the SA and NSA series in October 2024, then apply the ratio to the interpolated SA value in October 2025.) This approximation is good enough for aggregate series, but the underlying components might have issues.
The data hole is going to confront economists far into the future, and act as reminder of the fragility of state capacity when it is trashed by vandals.
Inflation-Linked Bonds
Cash flows associated with U.S. inflation-linked bonds (TIPS, or TIIS) are multiplied by a daily CPI index ratio. This is calculated based on a daily CPI index based on interpolating monthly CPI values.
The monthly CPI index level is associated with the first calendar day of the month three months after the data month of the report. That is, the October 2025 CPI value is associated with the January 1 2026.
For the days in the middle of the month — e.g., December 2nd to December 31 — the value of the daily CPI index is the linear interpolation between the September CPI value that is associated with December 1st.
(As an aside, this means that the carry associated with a TIPS is really a 2-month lag versus the data. If there were a strange spike in October, the daily inflation inflation rate spikes in December, showing up on December 2nd to January 1st.)
This means that a strict interpretation of the calculation would imply that the daily index would be NaN for December 2nd 2025 to January 30, 2026.
I have no idea what the fallback calculation is for TIPS in the case of missing data. People worried about a great many things, but most of the worries were about the government doctoring data — not stopping publication completely.
The fallback does not really matter unless there are cash flows from the government on the affected dates. If the government is making a payment (particularly a principal payment), the index ratio determines the cash flow. If anyone is going to sue, it would be about such payments. For secondary market trading, the index ratio matters, but all that matters is that both sides agree to the fallback calculation. If you do not like the fallback index ratio, you just adjust your pricing.
Although this disruption will add to the “TIPS risk factors” discussions, this is not too big a deal (unless you are involved with lawsuits about principal payments). The real problem is that the next time, the data might not come back, or might be replaced by a survey of burrito taxi pricing or “AI estimates.”
Americans in finance loved touting the advantages of the rule of the law in American capital markets, hopefully the rest of us will not have to sit through those lectures much longer.
No comments:
Post a Comment
Note: Posts are manually moderated, with a varying delay. Some disappear.
The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.
Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.