Hello: In late August 2006, the US Census Bureau (CB) released the 2005 American Community Survey (ACS) data on population, employed persons, journey-to-work, and many other demographic and employment characteristics such as income, poverty, employment status, occupation, and industry. These data are the first from the full implementation of ACS. This email reviews 2005 ACS data on population and journey-to-work for the Delaware Valley region which consists of five counties in Pennsylvania and four in New Jersey. Population Estimates The 2005 ACS population estimates are limited to household population only; they exclude group quarters population. To obtain 2005 ACS population, the CB estimated group quarters population and subtracted it from the 2005 total population estimates produced as part of the CB's Population Estimates Program (PEP). A comparison of 2005 CB population estimates with those estimated by DVRPC shows that the CB estimates are very close to DVRPC estimates. The differences between the two sets of estimates for six counties are less than 2.5 percent. The CB estimates for two suburban counties are overestimated slightly and one (Philadelphia) is underestimated, compared to DVRPC's estimates. An analysis of the 2005 census estimates of group quarters population shows that none of the nine estimates is reasonable. Group quarters population are either overestimated or underestimated. For example, group quarters population for Camden County did not change even by one person between 2000 and 2005. The Montgomery County estimate decreased from 23,257 to 22,837 persons rather than increased in this growing county. Although the magnitude of group quarters population is generally small, it increased the margin of error in the ACS household population significantly. Also, the erroneous group quarters population increased the margin of error in the 2005 ACS economic characteristics, such as the number of workers and means of transportation to work. Affected the most are counties with large group quarters population such as Montgomery County and Philadelphia. Journey-to-Work A review of the 2005 ACS journey-to-work by means of transportation indicates that the margin of error in the estimates is very large. Also, the CB could not develop all estimates for Gloucester County, New Jersey (277,000 people) because the "Number of sample cases is too small." As is known in Statistics, the sampling error increases inversely with the size of the variable. Since most people drive to work, the margin of error in drive alone mode is small. The opposite is true for those who walk to work. Nine out of 17 means of transportation are either underestimated or overestimated in Philadelphia as the difference between 2005 and 2000 estimates is larger than 20 percent. There is no clear pattern in the 2005 estimates of workers by means of transportation, some increase and others decrease. Also, there is no logical reason for the increase or decrease in the number of commuters that use a particular travel mode. For example, the number of workers in the DVRPC region that commute in 3-person carpool increases while those using 4-person carpool decreases. Significant increases and decreases in the means of transportation to work resulted from using a small sample of households which does not include group quarters population. Average Travel Time The 2005 ACS commuter travel time in the region is 28.3 minutes. This value indicates that the average commuter time has declined slightly between 2000 and 2005 (28.6 vs 28.3). It is interesting to note that the 2004 ACS travel time increased between 2000 and 2004 (28.6 vs 29.1 minutes). Based on this information, some people may conclude incorrectly that commuters' travel time increased from 2000 to 2004 and deceased from 2004 to 2005. No conclusions can be drawn based on this information since the differences between numbers are small and within the margin of error in the data. Also, the 2000, 2004, and 2005 census data are not comparable. They are oranges, apples, and lemons. The sample sizes are different and ACS data exclude group quarters population. Although the 2005 ACS sample size is larger than the 2004 sample, it is still about 12 percent of that used in Census 2000. As I have stated last year after the release of 2004 ACS data, the errors in the ACS estimates are large and the data cannot be evaluated or corrected easily. There are no accurate data between decennial Censuses. As Chuck Parvis stated correctly in his analysis of 2005 ACS data, it is not possible to "Calculate whether or not there is a meaningful difference between 2000 and 2005....that's more or less a professional judgment call." In summary, the 2005 ACS data cannot be used for transportation planning. The CB simply connot produce quality data from the ACS program comparable to the long-form of Census 2000 due to large sampling and non-sampling errors. The ACS sample size is too small, the population universe is estimated, and the data collection procedures are not consistent with Census 2000. Thabet Zakaria Deputy Director, Technical Services Delaware Valley Regional Planning Commission 190 N. Independence Mall West Philadelphia, PA 19106 Phone: 215-238-2885 Email: tzakaria(a)dvrpc.org Fax: 215-592-9125