IS the Consumer Price Index (CPI) accurate?

So we have all been talking about the CPI, inflation and interest rates. If you had looked at CPI closely you would have noticed that in 2022, when rents were going through the roof (pun intended), their impact on the "Shelter" component of CPI was marginal. This is because even though "Shelter" amounts for about a third in the "merket basket" that goes into CPI calcs, Rents and Owner Equivalent Rent (OER) are lagging data points in CPI.

Owner's equivalent Rent (OER) makes up about 23.8% and Primary residence rent another 5.9% in the CPI calculation.

The Bureau of Labor Statistics (BLS) uses Rent data for rental properties and for owner occupied properties, the BLS computes an equivalent "rental" price for the owned residence to come up with its calculations. Rather than use the "value" of the home, the BLS wants to use a "price" for the the shelter the home provides to arrive at the "cost of shelter" for a home owner.

So how does the BLS get all this data - well, through old fashioned panels! The BLS collects this data for about 40,000 homes, divided into six panels, with data refresh on each panel every six months (e.g., the 1st panel is sampled in January and July, the 2nd in February and August and so on). The BLS then uses this data to extrapolate and arrive at the OER data for the owner occupied properties. There in lies the problem - you need really good national data to get to some directionally accurate model representation.

From this chart you can see the discrepancy between 2022-present. The Shelter component of CPI seems to be lagging compared to other industry-based rent indices.

This data is lagging because tenants stay in properties for 2-3 years on average. While rents go up marginally dueing lease renewals (typically annually), a reset only occurs when the current tenant moves out and the property is marketed to a new tenant. So the semi-annual panel update will not be an accurate reflection of current market estimates. For example, with 2.5 to 3 years tenacy term, only 15-20% of tenants will change every 6 months (and presumably 3-4% per month). So accuracy on rentals itself gets quite complicated. Extrapolating that erroneous data to OER exacerbates that problem.

At Roofstock , we track about 110M single family residential properties (most detached SFR with a few small multi-family like duplexes, triplexes and quadplexes). We track dozens of data points on each of these properties, including a Roofstock Rent Estimate (whether it's a rental property or owner occupied property). So in reality if the BLS wants to look at other metrics to form an opinion, Roofstock has very comprehensive data on rental comps.

While home prices impact rents in the long-run, a shortage a housing supply might drive up demand, but it's unlikely to move rents proportionally, especially if rents have already gone through recent adjustments.

Looking at the most recent CPI breakdown, you can see that 12-month Change in Shelter from Mar 2023-24 was 5.7%

However, based on the charts above we can see that the "market" cost of Shelter is lower than the CPI estimate. Does this mean that the inflation data is actually not as bad but it might take us another 6 months to figure that out?

Would love to get a perspective from researchers and economists. Paul Briggs, CRE , any thoughts?