On September 8, 2006, I posted on this network, some comments on the
2005 ACS data on population and journey-to-work estimates for the
Delaware Valley region which consists of five counties in Pennsylvania
and four in New Jersey. In this email, I want to discuss the Delaware
Valley Regional Planning Commission's (DVRPC) experience with the 2005
ACS employed persons by place of work at the county level.
DVRPC uses census employed persons by place of work to develop the
number of jobs or employment. Employment data are used in the DVRPC
modeling process to estimate trip attractions by trip purpose. To
account for total employment, DVRPC factors upward census employed
persons by place of work to account for workers who have multiple jobs
and workers who were on vacation or sick during the survey days. DVRPC
has used this method successfully to develop employment estimates based
on the long-form of Census 2000.
The application of this method to 2005 ACS data on employed persons
produces erroneous results compared to those obtained from Census 2000.
The 2005 ACS employed persons by place of work results in reasonable
employment estimates for only three counties of the DVRPC nine counties.
Employment estimates for three of the remaining six counties are
significantly underestimated and three are overestimated. For example,
the employment estimate for Delaware County, PA is underestimated by
24,500 jobs or 10.3 percent compared to estimates from secondary
sources. Conversely, the employment of Burlington County, NJ is
overestimated by 16,300 jobs or 8.1 percent. These errors are expected
in the 2005 ACS results because of many reasons, including the ACS
sample size is too small, the population universe in estimated rather
than counted, and the data collection procedures are not comprehensive.
To produce reasonable employment data, DVRPC has used employment
estimates from other secondary sources such as BEA, BLS and ES 202.
Unfortunately, none of these sources can be used without adjustment.
BEA data have always been overestimated and BLS and ES 202 have always
been underestimated. Also, the margin of error in local employment
estimates is generally large. Hopefully, the planned accumulation of
ACS data for five years (2007-2011) may produce more accurate employment
estimates than the 2005 ACS data and data from secondary sources.
Deputy Director, Technical Services
Delaware Valley Regional Planning Commission
Philadelphia, PA 19106
Show replies by date