ACS Data Users Group

I'm comparing Google public data (http://bit.ly/1ibkhwL) vs. acf tabulated data.

Specifically, I'm looking at a big jump that appears between 2007 and 2008 in ACF data:
factfinder2.census.gov/.../0500000US06075
factfinder2.census.gov/.../0500000US06075

There's a jump from 764,976 to 808,976 that appears in ACF data, but not on Google data (799,185 to 808,001). From 2008 on, the data is very close, but not 2007 and below.

I'm trying to use the ACF PUMS data to break down SF demographics, and I'm stuck here trying to understand where this discrepancy came from.

FWIW, the numbers on the bay area census site agree with Google (www.bayareacensus.ca.gov/.../SanFranciscoCounty.htm)

It looks like the difference is between 5 year ACS and one-year ACS. You should figure out where the various sources are coming from.

Thanks for the quick response Matthew.

What's your recommendation if I want to look at more detailed ACS data breakdowns from 2005-on? Is it safe enough to just linearly scale the survey results up to a different population size?

This has to do with Population estimate vintages. The ACS number you reference is controlled to the population estimate for that year, from that year's vintage. So the 2007 ACS population is controlled to the 2007 vintage estimate of the population in San Francisco County on July 1, 2007. The 2008 estimate is controlled to the 2008 vintage estimate of the population in San Francisco County on July 1, 2008.

As a comparison, the google data appears to have pulled the 2009 vintage estimates for the population on July 1 for every year in the decade.

The difference is that one is the 2009 vintage estimate for 2007 and the other is the 2007 vintange estimate for 2007.

The different vintages can be found at:
www.census.gov/.../CO-EST2007-01-06.xls
www.census.gov/.../CO-EST2008-01-06.xls
www.census.gov/.../CO-EST2009-01-06.xls