First of all, they did not have data for most categories for October, but they did have data for some categories, so it wasn't completely "imputed". Second, they had to do something to estimate October data. This explanation published by BLS explains why they used the "carry-forward" method. It was really their best option. They admit that it probably results in an underestimation. Unfortunately, the media has mostly ignored BLS' explanations of the weaknesses of the data.
https://www.bls.gov/cpi/additional-resources/2025-federal-government-shutdown-impact-cpi.htm
6. Why did BLS choose carry-forward imputation?
BLS used the carry-forward method to impute data. BLS viewed retroactively collecting data for October as both costly and potentially inaccurate, especially since the CPI surveys are not based on respondent recall. For rents in particular, BLS collects data for an entirely different sample each month. BLS does not have data collection resources to collect 2 months of data in 2 weeks. Extending data collection for the October reference month through the end of November would have further delayed processing and release of November indexes and releases for subsequent months. In keeping with its mission, BLS publishes estimates as quickly as possible after the reference month. By using known and documented statistical methods, BLS was able to release November indexes within 10 days of the originally planned release date and is on schedule to release December data on the original timeline of January 13, 2026.
For October 2025, BLS used the existing, documented imputation algorithm in calculating the CPI. BLS did not deviate from existing, adopted methodology as described in the BLS Handbook of Methods for the CPI. Imputation is a hierarchical process for both uncollected prices in the Commodities and Services Pricing Survey and uncollected rents in the Housing Survey.
The imputation process first attempts to estimate a missing price or rent from the price change movements of ‘like-kind’ items observed in the current month. This is key, as the CPI measurement objective is to observe (or impute) prices as transacted strictly in the current month, not to look at past trends or to a forecast of future behavior. Only when there are no ‘like-kind’ items in the imputation source pool does the algorithm resolve to a carry-forward impute; that is, to proxy the unobserved price or rent as its last known collected price or rent. In October 2025, since all survey data went uncollected, the ‘like-kind’ imputation methods could not execute, so imputation resolved to the carry-forward method for all survey samples, including rent.
There are several reasons for using carry-forward rather than an alternative method such as a trend-adjusted naïve forecast, an interpolation method, or any other time-series or cross-sectional econometric model. First, BLS did not want to intervene and superimpose last minute, unvetted judgment into the process, as this could have been perceived as manipulating the data unscientifically. Second, BLS information technology was built and rigorously tested for the hierarchical imputation algorithm. Logistically, even if there were consensus agreement on an alternative imputation-of-last-resort method, BLS could not have written requirements, modified the systems, tested, and certified the changes in a timely fashion; that would have further delayed the release of the November 2025 CPI and subsequent releases considerably. Third, vetting a proposed alternative imputation method and gaining endorsement would take considerable time. A methodological change of that magnitude would require publication in the Federal Register and a public comment period prior to adoption.
Carry-forward imputation has specific effects depending on market conditions:
-In periods of general price increases, the carry-forward method may result in an underestimation.
-In periods of general price decreases, it may result in an overestimation.
-When most prices are not changing and some are rising and some are falling, the method reflects the majority behavior for uncollected items.
= new reply since forum marked as read