US DATA: CPI Imputation Shares At Least Improved In November
Dec-18 17:29
It doesn’t help shine any light on the wide range of questions around October data collection but the imputation percentages for the commodities & services survey at least improved in November after a latest deterioration in September.
Specifically, the different cell imputation share fell to 34% in Nov after a latest record high of 40% in September (data back to Jan 2019), 36% in Aug and 32% in Jul.
For context, this share started the year at 10% before climbing following budget and staff cuts. It also peaked at 15% in the pandemic when in-person surveys weren’t possible.
Different cell imputation is used when the BLS needs to expand the geographical area for a specific category.
With 66% using home cell imputation, 0% used carry forward as opposed to what was presumably a huge spike in October.
The housing survey imputation shares were similar to prior months however, with a non-interview share of 92% in both Nov and Sep.
For these non-interview cases: “rent is imputed by the average change in rent observed for housing units within the same nominal rent class (low, medium, or high rent) within the same location. If this source pool is insufficient, the pool is expanded across geography similar to the method used in the C&S survey.”