ACS 5-year estimates Census Tract Boundaries

For ACS 5-year estimates that straddle 2010 (2007-2011 to 2009-2014), or all census tract estimates from the 2010 Census Tract boundaries?  My guess would be that this would be the case starting with the 2006-2010 ACS estimates, but I just wanted to confirm. 

  • Yes, that's correct. The 2006-10 ACS used the 2010 tract boundaries and then so did each following ACS product, even up to the most currently available 2015-19.

  • Yes, sort of. Each 5-year ACS estimate uses the boundaries from the last year of its range (eg. 2007-2011 uses 2011). Tract boundaries from 2010 to 2019 are similar. There are a few certain specific tracts in upstate New York, Arizona, and Los Angeles whose boundaries or numbers changed in 2011 and 2012. Those changed tracts will not match up with data from non-matching years, so you'll see null data on your map. The most fastidious thing to do would be to use the correct boundary for each ACS estimate. That may be overkill, so next best would be to use 2011 for 2007-2011, and 2012 for years beyond that. You could use 2012 boundaries for all your data, and have a few nulls in 2011, or 2011 boundaries and have a few nulls in 2012-2014; depending what you're doing and where you're looking at, these changes might not be of concern to you.

    Also just to note that other boundaries, like CBSAs, places, county subdivisions, and a few counties, have more changes throughout the decade.

  • I think you're asking: did ACS products make a product-wide switch to 2010 Tract boundaries in 2010? 
    The answer is yes. 
    Census Bureau ACSO retains the individual household-level responses collected thru ACS. The responses are geocoded to an X, Y point location. And those points can be identified with new tract memberships or new city memberships (if boundaries change) during the 5 years that those records are used in ACS "5-year" tabulations. 
    That make sense?

  • Hi, Bernie.  Thanks for the information.  I had assumed that Census Tract boundaries were stable between Census years, but it appears as though this is not the case.   We are just focusing on New York City, so we may be okay.  Will have to check if there were any changes to NYC Census Tract boundaries

  • The changes in New York State are all in Madison and Oneida counties (except combining some water tracts off Staten Island), AND they all happened in 2011, so I don't think you have to worry about changes in NYC. Here's info on that:

    And though it's not relevant to you, for completeness, here's info on the Arizona and California tracts changed in 2012:

  • Hi, Todd.  Thank you so much for your help.   I'm trying to create an archive for my  education research organization with ACS data at the tract level on a small subset of variables.  Our primary focus would be NYC and I was planning on downloading 5-yr tabulations starting with the 2006-2010 through the 2015-2019 datasets.  Our organization primarily uses SAS so I was going to save these files as SAS datasets.  We have census tract data for students in NYC (the individual students are de-identified), so researchers could merge these tabulations with the student-level data.  The census tract variable on the student-level data correspond to the 2010 boundaries, so I wanted to ensure that ACS estimates for the years I'm using also corresponded the 2010 boundaries.  

  • There were some changes in 2011 and 2012 and then they stopped.  They do not change boundaries between Censuses, and the CA change was an error correction.  The 2020 blocks are way better than the 2010 blocks.  There were no changes in NYC tracts after the 2010 Census, and the 2020 tracts are very similar to the 2010.  They do change the political and administrative boundaries in between Census years, and do it more or less every December.  The ACS generally lags about one year.  There was a change in upstate NY.