Last week, two more comments on ACS 2004 data were posted on this network. The first is from Ken Cervenka who stated that the concept of “continuous surveying” appears very sound. I completely agree, we all love to have current data for transportation planning. Unfortunately, the proposed ACS program will not provide us with accurate estimates that we can use. Based on DVRPC analysis, the sampling and nonsampling errors in the ACS data are very large, and the CB simply cannot produce TAZ data comparable to those obtained from the long-form of Census 2000. There are many reasons for this statement, including the ACS sample size is smaller than that of the long-form of the decennial census; it is almost impossible to improve the quality of various data sets to estimate the “moving five year average;” and the ACS sample does not include group quarters population. As you stated correctly, the ACS approach requires not only sufficient sample sizes, but good annual weighting factors at the sub-county level. Re-weighting ACS samples by sub-county estimated population will not improve the quality of ACS data greatly. The CB would argue correctly that sub-county population estimates have larger errors than those at he county level. Many of the variables used by the CB to estimate demographic data, such as immigration and vital statistics, are only available at the county level. In addition, bias in selecting the ACS samples, large non-response rates from certain population groups, and lack of large marketing programs to promote the ACS and enhance the quality of data collection and processing, are additional reasons for the large errors in the ACS data. Sam Granato wrote that there is no need to even think about “reviving the long-form” because Congress has not forgotten yet the outcry from bureaucrats and people in 2000. Apparently, he is not impressed by the quality of either the long-form or the ACS data, and wants to develop his models based on SF1 data and “secondary source” information. I suggested bringing back the long-form because of many reasons, including the following: 1. There are large errors in the ACS data and the data cannot be corrected easily or used for modeling and traffic analysis. There are no solid basis for correcting the ACS estimates, such as census population counts; 2. The errors in the long-form data are much smaller than those in the ACS data. The errors can be corrected by data users based on SF1, SF3, and other census data. For example, the DVRPC region contains 9 counties in Pennsylvania and New Jersey, 355 municipalities and 1912 traffic analysis zones. The evaluation of CTPP data indicated that the population of 9 municipalities was erroneous, the employment in 15 percent of the zones was not coded correctly, and only 1 zone was not located correctly. The data in Tables 3-03 through 3-07 in Part 3 of CTPP cannot be used, not because of major errors in the long-form data, but because of an arbitory decision by the CB’s Disclosure Review Board regarding rounding and disclosure threshold; 3. The cost of the long-form is lower than that of ACS. In the late 1990s, the CB stated that the cost of ACS is much lower than that of the long-form, but in May 2005, it stated in Irvine, CA that the cost of ACS is equal to the cost of the long-form data. Since the ACS sample size is about 70 percent of the long-form sample size, the ACS cost is actually about 30 percent higher; 4. According to the CB, Census 2000 was very successful because Congress was interested in accurate population counts, which required the development of several expensive marketing and technical programs to enhance methods of data collection and processing and quality control measures. Most of these programs are not being implemented in the ACS program; 5. Based on DVRPC experience, the errors in data from “secondary sources” are higher than those in the CTPP. Also, some variables we use in modeling are not even available. For example, the CTPP is the only source for employed persons by place of residence and place of work at the TAZ level. To collect demographic and employment data at the TAZ level will cost millions of dollars. The long-form data are free. Because of this, DVRPC has decided to depend on census data for regional and transportation planning studies. Although not perfect, the responses to the long-form are the best source available to us for various transportation studies; and 6. Many planners remember that the CB wanted to delete the long-form from Census 2000, but could not do it because of the outcry of some bureaucrats from FHWA, FTA, OMB, state DOTS, and MPOs. With the ACS Program, Congress is now getting the outcry from the people annually, and some members prefer to hear all of the outcrying in the decennial census year. I think that the CB is completely capable of producing quality data in Census 2010, including the data we need for the transportation planning. After all, the CB has been doing it for 40 years. In order to produce quality data in Census 2010, the CB should improve its nonsampling procedures, call back, or reinterview some of those who do not respond to the long-form questions, develop an accurate geocoding of the place of work addresses, continue its successful public relations and marketing programs, change rounding rules, and eliminate the disclosure threshold.