The NY Times just released a new set of nationwide census map depicting some of the recently released data from the 2010 Census. Mapping the 2010 U.S. Census compliments the mapping site that the Times previously posted, Mapping America: Every City, Every Block, which used data from the 2005-2009 American Community Survey.
With both datasets side by side we can see important differences between them. If we look at census tract 68 in New York County (the tract where Baruch College is located), according to the 2010 Census there are 7,614 residents and the two largest groups are white (71%) and Asian (14%). If we look at this same tract according to the 2005-2009 American Community Survey, the estimate is 5,501 residents of whom 80% are white and 8% are Asian. What’s going on here? How can they be so different?
The 2010 Census is a snapshot, a one time count of what the total population was on April 1, 2010. The ACS on the other hand is a moving image, a rolling average based on five years of sample data between 2005 and 2009. They’re not going to be the same as they cover different time periods and are calculated using different methods. Also, the NY Times and most other publications often fail to report the confidence interval and the margins of error for the ACS estimates – for the 2005-2009 ACS, the estimate for Tract 68 is 5,501 residents, plus or minus 459 residents at a 90% confidence level.
Estimates tend to be more reliable as the population being surveyed increases or the geographic area increases in size. So you’ll get better estimates for total number of people than you would for a particular group, and you’ll get better estimates for a county then you would a census tract.
Which dataset should you use? For most of our students, who are often more interested in socio-economic variables like income and education, you’ll have to stick with the ACS. If they’re looking for population data (age, race, gender) and those socio-economic variables as well, they should get all of them from the ACS and not mix the two datasets together. The 2010 Census is not going to include ANY of the socio-economic data; these questions were cut and the ACS is the only place for these variables. But, if you have patrons who are interested in population change from 2000 to 2010 or they’re only interested in demographic variables, then the 2010 Census is the place to go.
I’ve included some maps from both mapping sites below that show tract 68 – they display the same data (total population and race) in a different format (2005-2009 ACS is a dot density map that shows race, 2010 Census is a shaded area map that shows total population change).
