The BLS tweaked the methodology after letting it go haywire for two years. This whole fiasco should be a career-ender for the top of the BLS.
By Wolf Richter for WOLF STREET.
October was the first month in the Consumer Price Index without the monthly mega-push-down adjustments to the health insurance CPI, which started with the October CPI in 2022 and went for 12 months through September this year, thereby ridiculing actual health insurance expenses that continued to soar. It downward-distorted CPI, core CPI, and most of all, core Services CPI to an ever-increasing extent month after month for 12 months through September.
Conversely, in the 12 months through October 2022, the health insurance CPI increases had been overstated, but to a far smaller extent.
So in today’s CPI data — my discussion: Beneath the Skin of CPI Inflation — the Bureau of Labor Statistics tweaked the odious health insurance CPI metric in a few ways, as expected, and on a month-to-month basis, the health insurance CPI jumped 1.1%.
But this 1.1% jump came after 12 months in a row of month-to-month plunges of about 4% per month, ultimately a 37% collapse in the health insurance CPI in 12 months through September, that took the health insurance CPI back to where it had been in August 2018, even though health insurance expenses have skyrocketed. And it’s from this August 2018 basis that the health insurance CPI increased by 1.1%.
The increases going forward may get larger, going from +1.1% for October to perhaps +2% in November and +3% in December because the tweaked version of the index now includes “smoothening” (via a moving average) which delays the impact of the positive values on the current index (more on this in a moment).
The year-over-year change is still hugely distorted because the 1.1% increase in October was from the base that had been knocked back to August 2018 levels.
So today’s 1.1% month-to-month increase reduced the year-over-year collapse from -37% in September to -34% in October. Each month going forward, the year-over-year collapse will get smaller.
This chart shows the health insurance CPI as index values, which were knocked back to August 2018, and just ticked up a smidgen from there today – that little hook at the bottom, which was today’s 1.1% jump, after the 37% collapse. The BLS has turned the health insurance CPI into chickenshit:
On a year-over-year basis, the health insurance CPI is now down 34% (see the little hook at the bottom of the chart), after having been down 37% in September. Each month going forward, it will be down less on a year-over-year basis.
After the model blew up.
This ignominious fiasco of an important metric within the CPI data occurred because the model that the BLS used to estimate the health insurance CPI – the “retained earnings method” – after working reasonably well for years, blew up amid the distortions and money flows during the pandemic.
Rather than coming up with an alternative estimate right away, back in 2021 when these issues became apparent, the BLS let this fiasco run for two years, overestimating by a moderate amount health insurance inflation in 2022, and causing health insurance CPI to just collapse over the past 12 months through September 2023, back to 2018 levels.
So the BLS finally decided to tweak the methodology of the index, after letting the collapse run the entire 12 months. It tweaked the methodology based on recommendations of the National Academies of Science, Engineering, and Medicine, Committee on National Statistics (CNSTAT), it said.
The tweaked version: increases to get larger over the next few months.
But it only tweaked the catastrophic old version, and this new version still relies on the “retained earnings method” that exploded during the pandemic. The BLS explains its tweaked methodology here.
The two main tweaks are:
- The index is now “smoothened” (using a moving average)
- The retained-earnings data is included twice a year, rather than just once a year.
The smoothening (using a moving average) has the effect that the impact of the actual increases are delayed. The 1.1% month-to-month increase for October would then get larger for November, perhaps to +2%, and get larger again for December, perhaps to +3%, as the moving average drops the prior negative data points and picks up the new positive data points.
Nevertheless, the new version uses the chickenshit data through September as the base.
The BLS should have come up with a different way of estimating health insurance CPI starting in 2021 when it became apparent that it was going haywire. And then it should have adjusted backwards the entire data series, as it does when other adjustments are made. But no.
The metric remains a fiasco, having now contributed to a significant understatement of core CPI and even more so of core services CPI for the past 12 months, and going forward, on a year-over-year basis for the next 12 months.
All supervisory employees and top-level management at the BLS who had anything to do with the health insurance CPI since 2020, the people that tolerated the old version for so long and that approved the tweaks of the new version, should be fired. This kind of fiasco should be a career-ender for higher-level folks at the BLS.
And here is my more temperate discussion of the CPI inflation data released today: Beneath the Skin of CPI Inflation
Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the beer and iced-tea mug to find out how:
Would you like to be notified via email when WOLF STREET publishes a new article? Sign up here.