RE: [CTPP] [ctpp-news] Demographic Analysis for Block Groups - Decennial Census vs. ACS 5 Year Estimates

Dr. Bernardin: Let me make clear that, when talking about earnings for resident workers, I was referring to the 3 categories (such as "less than $1250/month) and NOT to so-called "point" data. So far as I know, LEHD does not provide point data for income. I was not trying to say that LEHD provided the same level of data as ACS; hence the "apples and oranges" comment. But, for the small town in question, ACS and LEHD showed opposite trends. Thus, ACS showed median income declining quite sharply from 2000 to the 2005-09 period (even when allowing for the differences between Census long-form and ACS data), while LEHD showed an equally pronounced drastic change in the other direction, with proportionately fewer households in the lowest earnings category and sharply more in the highest. Back to apples and oranges. In theory, they could both be accurate. After all, income and earnings are not the same thing. The LEHD data only refers to the working population, whereas ACS surveys everybody. But I'm writing a planning study for this small town, and trying to convey what really happened to the local economy during the years post-Census 2000. We planners cannot always wait until we know for sure - the planning study must be complete in one month. Based on evidence "on the ground," the LEHD trend makes more sense. A new high-tech employer just moved in nearby, and there are a couple of new subdivisions with housing of a higher quality than this town has seen before. Therefore it seems that LEHD is giving an accurate picture, while ACS is in error. This fits with the sampling size problems of ACS for a town of barely 2,200 people, and the fact that the data cover a span of years, whereas each LEHD finding is for a single year. I am not, however, claiming to have the answers to these questions. And I have found ACS useful at larger levels of geography, where sampling error is smaller. Jonathan Lupton AICP Research Planner Metroplan Little Rock -----Original Message----- From: ctpp-news-bounces(a)chrispy.net [mailto:ctpp-news-bounces@chrispy.net] On Behalf Of Bernardin, Jr., Vincent Sent: Wednesday, April 06, 2011 6:20 PM To: ctpp-news(a)chrispy.net Subject: RE: [CTPP] [ctpp-news] Demographic Analysis for Block Groups - Decennial Census vs. ACS 5 Year Estimates This raises a really interesting question about what data source should be used for income (or earnings) at low levels of geography. There are applications, such as travel demand and land use modeling, which require household income data at very low levels of geography (e.g., traffic analysis zones). So, regardless of the error of the estimate, someone will have to wager their 'best guess' at income levels for small geographies for these endeavors. I believe what I would like to ask is, given this data need, what would provide our best guess - among professionals on this listserv? In the past, Census long form data was used for this purpose. I suppose I personally had simply assumed up to this point that ACS 5 Year estimates at the block group level would be our best data source going forward. That is, I so assumed until the post below raised the question in my mind - does LEHD provide an alternative? - is it a better alternative than ACS? I have only used LEHD for estimates of employment by place of work (and I haven't personally done that recently), so you will have to excuse my ignorance but at what level of geography does LEHD (or On the Map) provide estimates of earnings by place of residence? Also, does it provide an actual estimate of earnings (or median earnings, etc.) or does it only provide the number of jobs by earning level (less than $1,250/month, etc.) in several very broad categories? If LEHD provides a point estimate of earnings by place of residence at the block group level (or similar or lower geography), is there any documentation or does anyone know the confidence interval/MOE for these estimates as compared to those for ACS income data at the block group level? In the absence of such formal information on the LEHD side, what are folks best guess as to the accuracy of LEHD earning vs. ACS income estimates at low levels of geography? How does the sample size advantage vs. the deliberate introduction of noise in the LEHD data pan out? Would others generally agree with Jonathan's conjecture that LEHD is really the superior data source for this purpose at this scale? On a related note, has anyone else compared ACS income vs. LEHD earnings data - at any level of geography? Can anyone shed light on how closely they correlate? Thanks for entertaining my curiosity! Vince Vince Bernardin, Jr., Ph.D. Chief of Transportation Modeling Bernardin, Lochmueller & Associates, Inc. -----Original Message----- From: ctpp-news-bounces(a)chrispy.net [mailto:ctpp-news-bounces@chrispy.net] On Behalf Of Jonathan Lupton Sent: Wednesday, April 06, 2011 4:44 PM To: ctpp-news(a)chrispy.net Subject: RE: [CTPP] [ctpp-news] Demographic Analysis for Block Groups - Decennial Census vs. ACS 5 Year Estimates Ms. Cogburn: I can't address all of your concerns, but I can say I'd be careful with the 5-yr ACS at the BG level. The sample size is just too small, and margins of error too great. I agree with Ms. Becker that the old long form data probably had large errors at the small-geography level too, but the nice thing is they didn't tell us the margins of error! I have found the ACS useful and reliable at larger areas of geography, but the 5-yr product has shown signs of inaccuracy in our region. For example, one of our towns with a population of 2,200 showed a median HH income trend that looked odd (sharply declining median household income from 2000 to the 2005-2009 period). It could be true, but the trend change is completely different from what LEHD's "On the Map" product showed for approximately the same time period (2003-2009), for the same town, on workforce earnings by place of residence. As I understand it, LEHD comes from a survey which, while not perfect, gets a sizeable chunk of the workforce, while ACS resident surveys leave lots of holes. Admittedly, there's an "apples and oranges" aspect to comparing ACS with LEHD, and median HH income with workforce earnings. I'd like to see more research on the respective advantages of ACS and LEHD. Summary: I'd treat the ACS five-year product with care for BGs until we have more experience with it. And, if you haven't already, you might want to fiddle with the LEHD's new version of "On the Map," it's got a high "wow" factor for quick place-based data analysis. Jonathan Lupton Research Planner Metroplan Little Rock AR -----Original Message----- From: ctpp-news-bounces(a)chrispy.net [mailto:ctpp-news-bounces@chrispy.net] On Behalf Of Cogburn, Megan S Sent: Wednesday, April 06, 2011 2:58 PM To: ctpp-news(a)chrispy.net Subject: [CTPP] [ctpp-news] Demographic Analysis for Block Groups - Decennial Census vs. ACS 5 Year Estimates My name is Megan Cogburn and I am a Community Planner with the North Carolina Department of Transportation. My group is a part of the Project Development and Environmental Analysis branch and we are responsible for completing all of the NEPA documentation for bridge and highway projects statewide. Specifically, we prepare technical reports assessing potential project impacts on the human environment from the local/urban planning perspective. We currently use decennial Census data for our demographic analyses, however we are thinking of switching to use American Community Survey data to make our reports more current and since the American Community Survey has now replaced the traditional decennial Census long-form. However, it has come to our attention that ACS data is only available at the block group level for 5 Year Estimates (and not annually). Moreover, certain tables that were available for the 2000 Census are not available in the ACS 5 Year Estimates (such as household type by relationship, sex by employment status, and poverty status). Another glaring issue is that ACS 5 Year Estimates were just released in 2010, so there is no previous data to make historical comparisons. So, my question for the listserv is how other organizations are moving forward given the discrepancies between the two datasets. My group is trying to figure out where to get missing data, how to make historical comparisons, and also the best way to retrieve ACS 5-Year Estimates. A huge drawback for us is that we use block group data for multiple variables and this is only available for the 5 year estimates. In order to retrieve summary file data for block groups you have to use an Excel macro retrieval file that takes an extremely long time, download a massive file from their FTP site, or use the not so user friendly Data Ferret platform. Please advise! Megan Cogburn, MCRP Community Planner | Human Environment Unit NCDOT Project Development & Environmental Analysis e: mscogburn(a)ncdot.gov p: (919) 707-6062 f: (919) 212-5785 Email correspondence to and from this sender is subject to the N.C. Public Records Law and may be disclosed to third parties. _______________________________________________ ctpp-news mailing list ctpp-news(a)chrispy.net http://www.chrispy.net/mailman/listinfo/ctpp-news _______________________________________________ ctpp-news mailing list ctpp-news(a)chrispy.net http://www.chrispy.net/mailman/listinfo/ctpp-news _______________________________________________ ctpp-news mailing list ctpp-news(a)chrispy.net http://www.chrispy.net/mailman/listinfo/ctpp-news