An official website of the United States government
The BLS has been conducting research on whether administrative trade data can be used in place of directly collected survey data for more homogeneous product areas in the Import and Export Price Indexes (MXPI). The administrative trade data are federal government records of trade transactions reported by importers and exporters. The data are processed for statistical purposes by the Census Bureau and the Bureau of Labor Statistics. The BLS uses the data in the sample selection process of the Import and Export Price Survey.
The research has validated the methodological soundness to use administrative trade data to replace directly collected survey data for calculating MXPI. The historical time series of import and export price changes, measured with unit value indexes (UVI) based on the administrative trade data, are a research product. BLS research products can be adopted for use in official published data series (e.g., the CPI new vehicles research index). As such, the UVI dataset comprises unit value indexes currently under consideration to replace directly collected survey data to calculate the official Import and Export Price Indexes (MXPI) in 2025. Moving forward, the BLS plans to codify selection of import and export UVI that are determined to be high quality to replace or expand publication of official price indexes currently calculated with directly collected survey data, without a break in series.
Using administrative trade data to calculate price indexes reduces response burden and allows for numerous improvements. The administrative data are collected by other government agencies from importers and exporters and provide a nearly complete record of all international trade shipments. The administrative data provide significantly more observations than are currently included in the MXPI, potentially allowing more detailed indexes to be published. In addition, the values in the administrative data allow for item weights to be updated more frequently, making a superlative index formula possible. This methodological improvement mitigates substitution biases associated with the Lowe formula currently used in the calculation of the MXPI.
The primary concern with the indexes calculated from trade transaction records prices is unit value bias. A unit value is a quantity-weighted average price of all products within a unit. Unit value bias is mismeasurement of price trends reflecting changes in product mix within a unit instead of price changes for the products in that unit. When considering administrative trade data as a source to calculate price indexes, the research challenge is to mitigate unit value bias that is introduced when price and price index movements are driven by changes in product mix rather than changes in product prices. To address unit value bias, new statistical methods have been applied to mitigate the bias and results have been focused on homogenous product areas where unit value bias is less of a concern. Only those unit value indexes in the research phase that show minimal unit value bias will replace the directly collected survey data source.
The administrative trade data for imports and exports are organized by Harmonized System product categories—the 10-digit Harmonized Tariff Schedule (HTS) number and 10-digit Schedule B number, respectively—and contain various transaction characteristics including price and quantity. Transaction characteristics are combined to group like-records into unique entry-level items. For each 5-digit BEA End Use product category, the transaction characteristics that specify the entry-level items are selected by following the Match Adjusted R-Squared Method in Chessa (2021). Monthly weighted unit values are calculated for each entry-level item with price and quantity information. These entry-level items are then aggregated to the 10-digit Harmonized System classification level with a superlative Tornqvist index formula. The weights used in the Tornqvist formula are the current month's weights and the average of the previous year's weights following the method in Vartia et al. (2021). The current month's price is compared to the average price for the previous year. These entry-level item weight and price data are annually chained to the previous year using the Tornqvist formula. These lower-level Tornqvist unit value indexes are then aggregated to upper-level price indexes for 5-digit BEA End Use products, using the current methodology for the official Import and Export Price Indexes.
The research unit value index data series for 68 import and 56 export 5-digit BEA End Use product categories cover the period January 2017 through February 2024. Unit value indexes presented here are selected from 138 and 130 import and export BEA End Use product categories, respectively. The unit value indexes in this research series are considered to have minimal unit value bias, show similarities to current official MXPI, and are being considered for adoption for official published data series. BLS currently publishes official MXPI for some of these 5-digit BEA End Use classifications in Table 1. Import price indexes for selected categories of goods and Table 2. Export price indexes for selected categories of goods and in the MXP Database. For each research unit value index that has a comparable official MXPI, the BLS series ID is listed for ease of comparison. The research import unit value indexes, introduced in September 2022, and export unit value indexes, introduced in September 2021, were revised in September 2022 and in March 2023. (The previous versions of the research data series covering the periods January 2012 through December 2018 and January 2012 through December 2021 are available upon request.)
View the import unit value indexes in Excel (updated September 27, 2024)
View the export unit value indexes in Excel (updated September 27, 2024)
Publications describing the unit value index methodology are listed here. Note that some methodologies have changed.
Chessa, A. 2021. "A Product Match Adjusted R Squared Method for Defining Products with Transaction Data." Journal of Official Statistics. 37(2): 411–432. DOI: https://doi.org/10.2478/jos-2021-0018.
Fast, D. and S.E. Fleck. 2022. “Unit Values for Import and Export Price Indexes – A Proof of Concept,” in Big Data for 21st Century Statistics. University of Chicago Press, edited by K.G. Abraham, R.S. Jarmin, B. Moyer & M.D. Shapiro. Available at: https://www.bls.gov/mxp/data/unit-values-import-export-price-indexes.pdf.
Fast, D. and S.E. Fleck. 2019. “Measuring Export Price Movements with Administrative Trade Data.” Working Paper 518. U.S. Bureau of Labor Statistics. June. Available at: https://www.bls.gov/osmr/research-papers/2019/pdf/ec190080.pdf.
Fast, D., S.E. Fleck and D. Smith. 2022. “Unit Value Indexes for Exports – New Developments Using Administrative Trade Data.” Journal of Official Statistics. Volume 38. Issue 1. March, pp. 83–106. DOI: https://doi.org/10.2478/jos-2022-0005.
Fleck, S. and S. Paben. 2023. “Case Study 5: BLS Import/Export Price Indexes Measured Utilizing Census Administrative Trade Data,” in Mirel LB, Singpurwalla D, Hoppe T, Liliedahl E, Schmitt R, Weber J. A Framework for Data Quality: Case Studies, FCSM-23-02 Data Quality Framework Implementation Subcommittee, Federal Committee on Statistical Methodology. October 2023. pp. 30–35. Available at: https://www.fcsm.gov/assets/files/docs/FCSM.23.02_DQ_case_studies_FINAL.pdf.
Vartia, Y., A. Suoperä, K. Nieminen and S. Montonen. 2021. “Circular Error in Price Index Numbers Based on Scanner Data.” Available at: https://ssrn.com/abstract=3801530.
For questions pertaining to the unit value indexes or related research, please contact the International Price Program by email or by phone: 202-691-7101.
Last Modified Date: September 27, 2024