What data is used to calculate the CPI (Consumer Price Index) and if it is open, where can I find it?

2 Answers 2


The CPI is based on surveys conducted by BLS (Bureau of Labor Statistics). Those surveys collect price information which is reviewed by commodity experts. Those experts review (and possible adjust) the data based on their knowledge of the particular commodity. That adjusted data is then averaged to calculate the CPI. You can definitely get the values that are averaged in the final calculation. I'm less clear on how to find the actual survey responses.


Everything dvogel wrote is correct. Note that on the ftp page there are several datasets (including "ap" and "cu") which are labelled "consumer price index."

The Bureau of Labor Statistics (BLS) samples posted prices at various US businesses on a monthly basis. This sampling includes roughly 85 thousand observations per month. The BLS then creates and publishes price indices broken down by the type of good and the geographic location where it is sold. The original data are not made available to the public, but researchers can apply for restricted on-site access. See (http://ideas.repec.org/a/ucp/jpolec/v112y2004i5p947-985.html). The underlying dataset is referred to as the "CPI-RDB" (Research Database).

The BLS does not publish documentation for the underlying dataset-- their website does not mention the CPI-RDB or the option of applying for access. The best documentation I am aware of was written by an intern circa 2001-2005 named Teague Ruder, now at RAND (http://www.teagueruder.com/portfolio/projects/RDB.pdf).

Final notes:

1) Be careful when using data from the BLS ftp page. Many of the datasets contain typos or missing tab delimiters. As important, the geographic codes are not consistent between datasets.

2) The Billion Prices Project at MIT is an attempt to recreate the CPI data by scraping online retail. Some of their data may be available.

  • Wow, thank you for that additional information. I find it especially interesting that the available datasets "contain typos or missing tab delimiters". Really!? Is this the raw data that the government has used or is this an export after their calculations have been made. Fascinating. I'd like to think their data is accurate when they analyze it but you never know. (And welcome to StackExchange.)
    – RLH
    Jul 3, 2013 at 11:50
  • 1
    I don't have on hand an example of the missing tab delimiters or carriage returns-- I should acknowledge the possibility that I may have made a coding error at some point. The bigger problem is in matching geographic entities and other headers, e.g. "Dekalb County" vs "De Kalb County". The BLS also uses MSA definitions that do not match those of the OMB and Census. In most cases it is easier to access the data via FRED at the St Louis Fed.
    – user906
    Jul 3, 2013 at 16:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.