Friday, March 28, 2014

What A Difference A Year (or Two) Makes When it Comes to Population Estimates

Each year, the Census Bureau's Population Estimates Program (PEP) utilizes current data on births, deaths, and migration to calculate population change since the most recent decennial census and produce a time series of estimates of population. The annual time series of estimates begins with the most recent decennial census data and extends to the vintage (i.e. most recent) year.

Each "vintage" estimate includes revamped estimates for all the prior "vintage" years. With the release of the 2013 estimates, for example, the Census Bureau also releases estimates for each of the years from 2013 back to the most recent census, done in 2010. So there are new estimates for 2013, 2012, and 2011. When the 2014 estimates are released, the Census Bureau will release "readjusted estimates" (my words not theirs) for those previously produced estimates done in 2013, 2012, and 2011. Basically, each time the estimates are released, all the prior years get superseded by a new set of estimates.

This constant revision of the previous released estimates leads me to the statement, "What a difference a year makes!"

Below is a chart of the estimates (in dotted blue) and the Census counts (in red) since 1998 for what the Economic Development offices typically refer to as the "Mohawk Valley" - namely the counties of Oneida, Herkimer, Fulton, Montgomery, Schoharie and Otsego. These estimates are based on the ninth "vintage" year of the intercensal estimates program - in other words, the estimates for 2001, 2002, 2003, etc., through to 2009 are all from the estimates as they were released for vintage year 2009.

In 2010, a new set of estimates were released, adjusted in light of the Census 2010 numbers. These vintage 2010 estimates go all the way back and replace previously released estimates to the year 2001.

I wanted to compare the estimates, as they stood originally for 2009 with what the decennial census counts showed in 2000 and 2010.  Note that the 1999 number is from the vintage 1999 estimates, and how it compares to the Census 2000 counts.

Click to Enlarge
This comparison, I think, allows the viewer to see the relationship of the estimates to the reality of the census counts. In each case, the ninth year of the intercensal estimates is substantially lower than what the following years (2000 and 2010) Census counts find. The question becomes is this a systemic issue, and does it represent enough of a difference that we should be concerned?

In the end, if the estimates seem to be habitually lower than what the decennial Census counts show us, how much faith should we have in the estimates we saw recently released? More on this in a future post !