Submitted by Joe Carson, former chief economist at Alliance Bernstein
Policymakers and analysts involved in the lively debate on the future path of inflation need to consider whether the government statistical agencies have the tools or information to provide an accurate general inflation assessment. According to the Bureau of Labor Statistics (BLS), half of the data comprised in the consumer price index (CPI) was "imputed" in the past year.
Reported inflation can be whatever you want it to be. Still, it needs to be measuring what policymakers believe it is for it to be appropriate as a monetary policy tool. Price mismeasurement is a policy problem and perhaps soon a credibility problem for policymakers, as "missing prices" make inflation-targeting a meaningless policy tool.
CPI - "Missing Prices'
Since the pandemic, the standard practice of personal visits, which historically accounted for three-fourths of price quotes, was temporarily discontinued. Instead, price data was obtained entirely from online sources or through telephone interviews.
According to BLS, the change in data gathering practices has significantly lowered consumer price response rates. For example, the scale of uncollected prices for non-shelter goods and services ran between twenty and thirty-five percent in the past year, more than twice the average. Shelter prices for homeowners, which account for one-fourth of the price index, are regularly "imputed" each month. Taken together, that means price "imputations" and not actual transaction prices have accounted for more than half of the CPI index for the past year.
Price imputations have always been a controversial issue. Government statisticians have used "imputed" prices for things like owner housing since it was conceptually estimating a cost-of-living index and not a standard price index. However, the inclusion of "price imputations" creates ambiguity and subjectivity, lessening its use as a market-based inflation index and a policy tool.
The pandemic has magnified the issue of price imputations. The measurement of owner housing has become more absurd. In the past year, all of the rent data was collected by telephone, far above the two-thirds average. Also, roughly thirty-five percent of rents every month were uncollected. While that sounds high, and it is, before the pandemic uncollected rent response rates ran consistently in the high twenties.
During the pandemic, a record number of rents were unpaid. Still, data collectors classified due rents as fully paid if the landlord "expect payment in full, regardless of when."
With no proof of current or future payment, word-of-mouth rents are included in the CPI to estimate primary residence rents and implied rents for owner housing. But house prices based on actual transactions, with proof of payment, are not. From an economic and statistical standpoint, why isn't this called-out to be what it is, "price-construction-trickery. "
The monetary policy oxymoron; price stability drives its decisions but reported price indexes do not offer a stable or steady flow of actual prices. They are replete with " missing" and "imputed" prices.
In the past few decades, changes in measurement practices coupled with "missing prices" have resulted in the "noise" of inflation being louder than the "signal" (i.e., reported inflation). Policies that follow the noise can preempt inflation cycles and awful economic outcomes, but targeting the signal can't. During the housing bubble, the Fed ignored the "noise" and followed the "signal." Will history repeat?