If the question’s whether we can trust the inflation data, the answer is, was and forever will be “no.”

One of the longest-standing macro jokes in the universe maligns the concept of “core” inflation as a measure which accounts for everything except the stuff people actually need, namely food and fuel.

More recently, the Fed turned its focus to an even narrower measure that strips out goods and housing, leaving “supercore” inflation, a metric which implicitly pretends that as long as prices for things like haircuts are stable, all’s well in the world.

Even if you set aside the kind of sarcastic derision associated with jokes about inflation metrics that don’t account for food, shelter and energy, there are methodological critiques which suggest that from a purely academic perspective, we just don’t do a very good job of capturing price trends, even when we give an honest effort.

But those questions nod to a different kind of “trust” than that which is violated when the government sets about deliberately underreporting inflation on a consistent basis in order to cover up a cost-of-living spiral that might jeopardize the political fortunes of the executive. That’s the kind of thing which goes on in Turkey, and there are questions as to whether it might soon happen at the BLS, under Donald Trump.

When a reader asked about such concerns on Wednesday, I noted that at least initially, the concern isn’t so much about a systematic effort to rig the data, but rather about the read-across from Trump’s bureaucracy purge. “Data mis-measurement in the CPI/PPI series is more likely to be a function of staff cuts which constrain the BLS’s capacity to cast a wide net than it is malicious intent,” I wrote.

With that in mind, have a look at the chart below (hat tip to Bloomberg’s Odd Lots column for surfacing this in the mainstream financial media on Wednesday), which shows the breakdown of imputed prices in the CPI release by month.

As the BLS explains, “home cell imputation” entails guesstimating (they don’t use that word, obviously) unobservable prices using collected prices “within the same item and geographic area.” By contrast, “different cell imputation” involves guesstimating uncollected prices by way of collected prices “of the same item in other geographic areas or from collected prices of related item categories in the same geographic area” (emphasis mine).

As the figure shows, the share of uncollected prices conjured using different cell imputation was 35% in Tuesday’s CPI release, up nearly fourfold since Trump took office.

Is that a problem? Well, maybe. It’s not ideal, is it? If the goal’s to accurately measure prices, the best way to do that is to observe them (the prices) directly. If that’s not possible, then surmising them by way of prices for the same item in the same region’s the next best thing. Once you start imputing prices based on the same item sold elsewhere and/or based on items that are related to the item you’re trying to measure but aren’t quite that item exactly, precision will suffer pretty much by definition.

As Bloomberg emphasized in the linked article, the recent spike in different-cell imputation at the BLS is directly attributable to Trump’s federal hiring freeze.

“Fewer BLS workers means fewer people available to conduct surveys and actually go out and directly observe prices, hence the need for imputation,” Tracy Alloway wrote, on the way to noting that although the trend isn’t “exactly an unexpected development, the speed at which it’s happening is stunning.”

You could say the same thing about any and all “developments” to do with “Trump 2.0,” including his generalized push to consolidate power in the executive.