A Better AEI Graphic of Inflation Over the Past 20 Years
What happens when we take the AEI graphic of items that have had high and low inflation, but extend it to all categories?
What happens to a take deferred? I did most of this take two weeks ago, when everyone was talking about it, before being asked to do (and having to prepare) House subcommittee testimony on inflation (check it out here) and then watching the whole banking system explode. But maybe you want a break from the ongoing financial crisis, so I’m posting it now.
Thanks for reading Rortybomb! Subscribe for free to receive new posts and support my work.
This graphic is in the news again:
Its creator is Mark Perry of the American Enterprise Institute, who last posted an update to it in July 2022. He’s been doing a version since at least 2016, and if you read enough economics blogs or content you’ve probably seen some iteration of it.
People are talking about it again after Marc Andreessen posted it under the headline “Why AI Won't Cause Unemployment.” Andreessen describes what people generally take away from it - blue line capitalism and dynamic, red line government regulations and stagnant:
The lines in blue are the sectors where technological innovation is allowed to push down prices while increasing quality. The lines in red are the sectors where technological innovation is not permitted to push down prices; in fact, the prices of education, health care, and housing as well as anything provided or controlled by the government are going to the moon, even as those sectors are technologically stagnant. [...]
The sectors in red are heavily regulated and controlled and bottlenecked by the government and by those industries themselves. Those industries are monopolies, oligopolies, and cartels, with extensive formal government regulation as well as regulatory capture [...] Whereas the sectors in blue are less regulated, technology whips through them, pushing down prices and raising quality every year.
Matt Yglesias noted on twitter that he’s “come to think it's misleading — by being very selective in which categories of labor-intensive services it chooses to chart, it's generated a narrative that relative price shifts are just about government regulation.”
That seems correct to me; these categories are pretty loaded. Let’s see if we can do better by including every possible category.
A better graphic
First, let’s download all of the current Consumer Price Index (CPI) data off the BLS download site, which is stored in a relational set of text files, and merge them together. (The code is here, you can follow along if you’d like.)
Here we have a lot of categories for CPI inflation. Unfortunately the data files don’t have the lowest item within a category. For instance there’s data for “Appliances,” “Major appliances,” and “Other appliances” - but no way within the data (that I can find) to know that “Appliances” is a higher-level category containing the latter two. We want to get as granular as possible, and we don’t want to double count in the presentation of price increases (which would happen if we showed all three).
So I manually go through Table 2, CPI by detailed expenditure category and make a list of what is the lowest level with data present.I also only look at what the BLS categorizes as “goods” and “services,” excluding energy, food, and the BLS’s own indexes (e.g. “all items”).
Since the BLS is constantly changing categories, we have to select the items that exist in both January 2000 and February 2023 to duplicate the chart. That leaves us with 62 categories. Doing a quick glance (and seeing in Perry’s own chart) the year-by-year evolution over time doesn’t really tell us much, so we can go with a simple bar chart for overall change. Let’s chart that here in full:
(You can get a csv file of this chart here. If you want to go wild, you can see the increases for all potential categories and indexes, not just the lowest levels, since 2012, when many new items were introduced, as a csv file here.)
There are a few key takeaways looking at it this way:
In our version of the AEI chart the number one item isn’t health care but ‘delivery services,’ which is “fees for delivery of items such as letters, documents, and packages at non-US Postal Services facilities.” Think UPS or FedEx. This is pretty far from a government monopoly, indeed it’s the private sector alternative to a government program.But it is services and it is labor intensive.
The biggest thing, to me, isn’t “regulations” but whether it’s a service or a good. This is academically well known, check out The Missing Inflation Puzzle: The Role of the Wage-Price Pass-Through, which finds that the low inflation of pre-pandemic era “can be traced to a growing disconnect between unemployment and core goods inflation” and that “increased import competition and rising market concentration reduce pass-through from wages to prices” when it comes to goods.
This question - why are goods so low? - is one potential reason academics wonder if monetary policy is losing its potency in stabilizing demand. See Structural Changes in Investment and the Waning Power of Monetary Policy by Justin Bloesch and Jacob P. Weber for more on this. What if the business cycle is stuck between a zero lower bound to heat up demand and the Fed needing to hike so much people don’t go to restaurants to cool demand?
Mileage varies, but I see ‘motor vehicle repair,’ ‘day care,’ and ‘alcoholic beverages away from home’ all in the same higher end ballpark. These three categories are subject to very different kinds of regulations, yet they all have the same price increases. What they share in common is that are all labor intensive.
I bought a TV for $300 in 2000; a TV now doesn’t cost $6 (as indicated with 98 percent deflation). It’s just that it’s a lot nicer and the numbers adjust for that. But that’s not going to be true for all goods. Tires, for instance, have noteworthy positive inflation even with the same kinds of global supply chains.
It’s funny to see AEI saying that inflation, over the long-term, is a function of market structure and regulations, not unlike the ‘greedflation’ argument. Though, as Seth Ackerman at Jacobin, channeling Keynes and Robinson, notes, “the level of wages is the most important single determinant of both the supply of and the demand for goods and services.” And goods are a part of the economy where wages are less determinative of final prices.
Just for a visual version of these points, here’s wages against goods (“commodities”) and services inflation. There’s little-to-no relationship on the goods side, but a strong one on services:
So there you have it. Stepping back and looking at all the categories tells us a much different picture. Don’t ever give up on your take, even if it gets cold.
I had done this anyway so I could do some distributional analysis of price increases, including duplicating this work by Governor Waller. If there’s a bug, it would be here.
I do make one change though; I include “Full service meals and snacks” (i.e. restaurants) as a service, though BLS includes it as a food. (The BEA, when calculating PCE inflation, considers it a service, and whether to count it is part of the recent debate over the divergence of CPI and PCE non-housing services inflation. I’ve started including it with an asterisk in my own monthly CPI analysis.
“Postal” wasn't introduced as its own category until later, “Postal and delivery services” exists in 2000 but isn’t a lowest level category; both are much slower than “delivery services.” Note that, since this category is consumer facing, it excludes “Business related services and charges for the delivery of freight such as household goods and automobiles. Shipping and handling charges of online purchases paid to non-delivery service outlets.”
Great post on how there's more than meets the eye on this chart.
I for one would be very interested in a Youtube tutorial! Most familiar with Python myself, but a walkthrough of what you did in R would also be interesting.
I use Mark's chart to suggest lack of innovation. (In a Paul Romer kind of way) That could come from regulatory barriers, investment or ideas. What data set reveals which? Can we map that covariance?