What Happens if CPI Isn’t Released?

One thing I’ve stopped worrying very much about is a government shutdown. It could even be a good thing, given the bloated deficit, except for the fact that the government basically keeps spending anyway. The federal government employs about 4.5mm workers, and no more than 800k have every been furloughed – moreover, many of those furloughed workers often receive back pay. Social Security gets paid, Treasuries get paid, and the wheel keeps turning. That’s not a guarantee, of course – it’s possible that an extended shutdown could cause Treasuries interest to not be paid, but we all know that before that happens, the Fed would just print the money and make sure the checks go out. At worst, there could be a one-day technical default, if important people had given the heads-up to insiders to get really long CDS.

But my cynicism is getting the better of me so let’s turn to what could happen in a shutdown that impacts the inflation markets: in the past, some data releases of federal agencies have been delayed (or their quality impacted), and if the delay was long enough then it could affect TIPS. Lots of people are asking about this, so I thought I’d lay out what would happen and how.

First of all, the quality of the CPI data could potentially be impacted. That has happened in the past, because data collection agents are not ‘essential workers’ so if the government shuts down, a lot of the data collection stops. This is less of a problem than it has been in the past, though, because a lot more of the data is collected electronically than in the past. For example, the new cars sample is no longer collected by hand but is sourced from J.D. Power. Prescription drugs data is partly supplied by one large firm that didn’t want to allow data collectors to collect data in store. A similar story applies to apparel. Many of these ‘big data’ changes are discussed in this BLS white paper, but the point is that these changes also mean that the quality of the data won’t be impacted as much as would be the case if data collection was entirely done by hand as it once was.

The bigger potential problem is that the CPI report could be delayed.[1] The NSA CPI is used almost exclusively as the index in inflation swaps, and is the index that determines escalation of TIPS principals. Other subindices are used in contract arrangements (for example, in long-term airplane purchase contracts), but those applications are generally less urgent.

If the BLS is unable to release the CPI on October 12^th, what happens? The first thing to know is that the September CPI (which is what is released in October) is only relevant to swap payments and TIPS accruals in November and December. For each day in November, the inflation index is interpolated between the August and September prints; for each day in December, the inflation index is interpolated between the September and October prints. Ergo, missing the September print would make it impossible to settle inflation swaps payments – but more importantly, every TIPS trade that settles in November or December would be impossible to settle because the invoice price couldn’t be calculated.

Fortunately, the Treasury thought about that a very long time ago. Title 31 of the Code of Federal Regulations (CFR) spells out what would happen if the BLS didn’t report a CPI by the end of October (it also spells out what happens if the BLS makes a large change to the CPI, or stops calculating it). In a nutshell, the Treasury would use the August CPI index, inflated by the decompounded year-over-year inflation rate from August 2022-August 2023:

CPI Calculation

I’ll do the math for you. If the CPI isn’t released, the figure for September will be 307.94834, which is +0.3004% on the month. While that sounds very convenient, since economists are forecasting a +0.3% m/m change for this data point, remember that the economists’ +0.3% is seasonally adjusted while the +0.3004% change is NSA. The difference is that 0.3004% NSA is about 0.50% SA this month.

Naturally, this wouldn’t matter very much in the long run; once the October CPI was released at the proper level the artificial change from Sep-Oct would wash out the artificial change for Aug-Sep.

Except, that is, for one pain-in-the-ass way, and that is the second part of the code snippet shown above: the Treasury would never adjust the official number back to match the BLS back-dated release of September CPI. Forever after, if you ran the sequence of monthly Treasury CPI Index numbers and the BLS CPI numbers, they would be exactly the same except for the one data point. The economic significance of that approaches zero, but the Inflation-Guy-Irritation figure on that approaches infinity.

So let’s hope cooler heads prevail.