This thread critiques Section I of Oren Cass's paper, "The Cost-of-Thriving Index." It includes 35 tweets. Eventually, the threads for each of the sections of his paper will be linked in a master thread. /1
Oren indicates Section I will be "a review of those [inflation measures] already available." (p7) However, he actually only reviews two measures, the Consumer Price Index for All Urban Consumers (CPI-U) & the Personal Consumption Expenditures Chain-Type Price Index (PCE) /2
He characterizes the CPI-U as the federal government's preferred inflation measure, but it's more accurate to say that for policy purposes, that has historically been the measure used. Since we're interested in empirics rather than politics, that's almost completely irrelevant /3
The major federal agencies examining income and wages began abandoning the CPI-U in the late 1980s because it suffered from significant flaws (see my 2015 essay: forbes.com/sites/scottwin…). By 1993, it had been downgraded in favor of the CPI-U-X1 by BLS, Census Bureau, and CBO. /4
The CPI-U-X1 didn't fix all of the flaws in the CPI, and by 1999, a new measure, the CPI-U-RS (now CPI-U-XRS) was introduced, and estimates were created back to 1978. This is the measure that BLS and the Census Bureau tend to use today. /5
But Oren is focused on the CPI-U, which to this day retains for the estimates from December 1978 to December 1998 all of the flaws corrected by the CPI-U-XRS. Keep that in mind. /6
Oren notes that the Fed switched to the PCE in 2000. CBO also switched to the PCE in 2012 for its income analyses. An unambiguously better feature of the PCE relative to the CPI-U is that it account for both "upper level" and "lower level" substitution bias (back to 1929). /7
Upper level substitution is when people switch from buying oranges to buying apples when their relative prices change. Lower level substitution is when they switch from buying red delicious apples to Macintosh apples. /8
Accounting for the ability of consumers to substitute is important because consumers have options when prices change--they dont have to keep buying what they used to, they can rebalance & avoid the full utility loss they wld endure absent this ability when some prices rise /9
The CPI accounts only for lower-level substitution, and only since 1999. Consequently, it shows higher inflation over time than the PCE and when incomes are adjusted using it, the increase is smaller over time than when the PCE is used. /10
(I'll note in passing that Oren claims the CPI does not account for substitution at all, but that is not true--it does account for lower-level substitution over the years he is comparing the two indices in this section.) /11
Oren downplays this advantage of the PCE deflator by quoting researchers who say it makes only a small difference and by displaying a chart purporting to show it only makes a small difference. /12
The first thing to note is that it actually could be said to account for over half of the difference, based on Oren's own chart. Second, all the evidence he cites in support of his claim come from 2002 data or later. /13
That's important because Oren will subsequently show affordability trends back to 1985. Most debate about trends in living standards relies on data going back to 1979 or 1973. /14
Whatever difference substitution makes from 2002 to today in how the CPI and PCE differ, it makes a bigger difference prior to 1999 when the CPI had accounted for NO substitution. Oren's discussion minimizes this clear superiority of the PCE by not mentioning any of this. /15
Oren highlights, instead, the different weights that housing & medical care receive in the CPI & PCE. He discusses which weights are right--you get the impression that he thinks the CPI weights housing better and PCE medical care better and the weights should sum to >100% /16
In the end, he adjusts the PCE upwards to account for the higher inflation caused by the bigger housing component in the CPI, but as @MarcGoldwein pointed out on Twitter, that should probably be offset by lower inflation caused by the bigger medical care component in the CPI. /17
At any rate, all of these issues over weighting are really a distraction, as I noted in my 2015 essay. BLS has a more recent inflation measure called the chained CPI that accounts for both upper- and lower-level substitution bias and weights things similarly to the CPI. /18
Guess what? It shows very similar inflation rates as the PCE--lower than the CPI, not higher, as Oren would have it. (Oren only mentions the chained CPI in a footnote.) From 2002-18 (the period Oren looks at) the chained CPI and the PCE both increase by ~34%, CPI-U by 39%. /19
Unfortunately, the chained CPI only goes back to 2000, so it is impractical for many income studies. But since it shows v similar inflation rates as the PCE, & because both fully account for substitution bias, the PCE shld clearly be preferred to the CPI for long-run analyses./20
What else to say? As noted, Oren argues the housing share of consumption is too low in the PCE, using a stylized example of a renter with $75,000 in income looking for a 3-bedroom unit. /21
With the weight given to housing, the renter would be able to afford such a unit at the 40th percentile of rents only in markets where just 1/6 of Americans reside. If this stylized example sounds confusing to you, it's not you. /22
Of course different kinds of households spend different shares of their budget on housing, and they spend correspondingly more or less on other goods and services. /23
At the extreme, homeowners without mortgages have minuscule shelter costs and experience housing "price increases" mainly as increases in home values. /24
There's a similar problem in his discussion of the college share of spending. Of course some households with children of college-going age will have bigger shares of their spending devoted to that category (for a few years). /25
No single price index can be anything but an average. It may be of interest to look at certain groups at certain life stages, but doing so doesn't negate all the complications of inflation measurement. /26
Oren is struck by the fact that Toyota Camrys have gotten more expensive since 1996, but he doesn't note that there are cars like the Chevy Spark that are cheaper today than a Camry was back then (before adjusting for inflation). /27
He says you can't buy a TV that sold for $500 in 1996 for $15 dollars today, though it didn't take me long to find a (manufacturer refurbished) 1997 Sony Trinitron (retail $2,500 in 1997) on eBay for $395 (a drop of 84%). /28
He says it's cold comfort that phones have gotten cheaper if you can't afford an iPhone. Of course, an iPhone replaces a lot more than just a phone (camera, GPS), so it eliminates other purchases that would have been made in the past. /29
(And iPhones are hardly the only cell phone available.) Mostly these are ineffective attempts to show that conventional inflation adjustment is somehow painting an inaccurate or incomplete picture of changes in living standards. /30
Finally, I'd be remiss in not mentioning that on p7, Oren presents a chart showing that the median weekly wages of full-time male workers have risen more than inflation since 1980 only for men with a bachelor's degree or higher. /31
As I've argued many times, showing wage trends by educational attainment is misleading because men are getting more and more education over time. That means that, eg, high school dropouts are a more disadvantaged group today than in 1980. /32
Therefore, their wages would go down simply because a more unusual group today is being compared to a less unusual group in 1980. The same is true at every level of education. /33
College graduates today are less "elite" than college graduates in 1980, so their wages today are lower than would be the case if it were rarer for workers to have graduated from college. /34
If you're interested in this mathematical phenomenon, google "Will Rogers Paradox". That's it for Section I /fin
