Brave New World Using the Why? ACS will provide • Better quality data • More frequently (every year) • Better knowledge of local areas • Cost savings ($1 billion) Goal is not to produce a population count but rather to produce the characteristics of the population. Effect on Decennial Census No more long form — In 2010, only 100% data Better Quality Data Over Decennial Because • Field representatives in every county will have local knowledge More highly trained, greater longevity, more highly paid How to reach hard to enumerate populations If a language barrier exists How to handle local situations • More extensive followup Computer-assisted telephone interviews as well as in-person visits However… Smaller sample size – 12.5% instead of 17% — 12.5% after 5 years; initially is about 2.5% per year — Oversampling in smallest governmental units so “about 2.5%” actually ranges from 1.7% to 10% — Bureau projects that “the estimates of sampling error for the five-year ACS estimates…will be about one-third higher than…decennial census estimates. We believe this is acceptable given the reductions of nonsampling error…” How can they do it every year? Rolling sample on a five-year cycle Huh? (Years reflect data collection) Release Schedule Data Available Now! • All States • Puerto Rico • Most areas of 65,000 or more What’s Different? • Data for smaller communities (<65k) in 3- and 5-year averages – Use larger communities’ 3- and 5-year averages to compare with smaller communities • Margin of error more transparent – All estimates presented with lower and upper bounds of 90% confidence interval – Can be very large compared to the estimate Median Value of Owner Occupied Units by Year Built Statistics Break! 32 23 25 23 25 24 25 23 23 23 23 Mode 24 24 23 18 18 18 32 23 32 23 Implications with Multi-year Averaged Data Green Handouts Estimates lag behind the actual trend • Handicap in beginning, especially • Fluctuating figures hard to track – will read as flat line Implications with Multi-year Averaged Data • What geographic boundary applies? – most recent within the average (will result in more TIGER updates) • Which year’s dollar value applies? – most recent within the average (others inflation-adjusted) • With no single Census date, what does “residence” mean? – ―current‖ instead of ―usual‖ (2 months rule) – seasonal populations may appear in any category of "second residence" (3 questions in 2005 questionnaire) Data Collection / Reference Period Griffin & Waite, 2006, p. 212 Summary - Comparisons with Decennial Data Yellow Handout At-a-glance Comparison See also, Blue Handouts Variable-Level Comparison Understanding the Numbers – Use the MOEs Comparing Two Estimates • If have two estimates, need to determine if the apparent differences are ―real‖ • Quick and dirty method is to ―eye ball‖ whether the confidence intervals overlap Interpreting Estimates Comparing Two Estimates (the easy way) • If the confidence intervals of two estimates do not overlap, then the two estimates are statistically different • If the confidence intervals of two estimates do overlap, then the two estimates are not statistically different (maybe) Sampling Error & Standard Error • Sampling error occurs when estimates are derived from a sample rather than a census (complete count) of the population. • Standard error is an estimate of sampling error – how precise the survey estimates are Sampling Error & Margin of Error • Margin of Error = standard error for a given confidence interval (typically 90 percent). A measure of the precision of the estimate at a given confidence interval • Sampling error is often reported as the estimate ―plus or minus‖ the margin of error Margin of Error (MOE) • MOE = 1.65 * Standard error 1.65 is used for the 90 percent confidence interval • Standard Error = MOE/1.65 Comparing Two Estimates • Need to do a formal test of statistical significance if the confidence intervals do overlap (The Hard Way) Statistical Testing - Steps 1. Calculate the difference in the estimates 2. Calculate the standard errors of each estimate Statistical Testing - Steps 3. Calculate the standard error of the difference 4. Calculate the MOE of the difference 5. Compare the difference between the estimates to the margin of error of the difference. Statistical Testing - Steps 6. If the difference in the estimates is greater than the margin of error of the difference, then you conclude that the two estimates are statistically different 7. If the difference in the estimates is less than the margin of error of the difference, you conclude that the two estimates are not statistically different. Go to Spreadsheet (The easy way to do the hard way) Two Tools http://www.mdp.state.md.us/msdc/SDC_meeting_ 040407/SDCmeeting040407.htm (little simpler) http://www.sdcbidc.iupui.edu/sharing/tools.html (little more flexible) Summary • 2005 ACS is HH pop and not total pop • Use CB estimates, NOT ACS for A/S/R/ state & county estimates • Need to evaluate apparent differences to see if statistically significant • A good idea to look at MOE to evaluate how good an estimate is • May need to wait for 3 or 5-year estimates to get data with acceptable margins of error Funding Issues • Fight every year but especially for ramp up (advocacy always welcome) – In same funding bill as Homeland Security – In FY2007, held to FY2006 budget • Affects planning and implementation – handheld data collection devices – Group Quarters behind schedule For More Information American Community Survey Web site Advanced methodology section especially (www.census.gov/acs/www) American Community Survey Web site State Data Center network for Using the Data and Advanced Methodology local areas sections especially (http://www.census.gov/sdc/www/) (www.census.gov/acs/www) Subscribe to the ACS Alert at: http://lists.census.gov/mailman/listinfo/acs-alert New Variables in ACS (2005) • Whether the household received food stamps in the previous 12 months and their value • The length of time and main reason for staying at the address (for example, permanent home, vacation home, to attend school or college), and • For women ages 15-50, whether they gave birth to any children in the past 12 months. Possible Changes (2008) New Variables • Marital history • Health insurance coverage • Veteran’s service-related disability Deleted Variables • The length of time and main reason for staying at the address Possible Changes Other Changes (2008) • To stay consistent with 2010 census • To improve response on some items Testing New Item – 2007 • Field of Bachelor’s degree Might be added in 2009 Table Number Trivia B = basic or base tables provide the most detailed estimates on all topics and for all geographies. C = collapsed version of a B table very similar to a B table with the same number (e.g., C07001 and B07001), but two or more lines from the B table have been collapsed to a single line in the C table. For example, the lines "75 to 79 years", "80 to 84 years" and "85 years and over" from a B table may be collapsed to a single line of "75 years and over" in a C table. * Not every B table has a collapsed version. Table Number Trivia The next two characters identify the primary subject of the table. 01 = Age and Sex 02 = Race 03 = Hispanic or Latino Origin 04 = Ancestry Etc. The next three digits are a sequential number, such as 001 or 002, to uniquely identify the table within a given subject. Table Number Trivia For select tables, an alphabetic suffix follows to indicate that a table is repeated for the nine major race and Hispanic or Latino groups: A = White Alone B = Black or African American Alone C = American Indian and Alaska Native Alone D = Asian Alone E = Native Hawaiian and Other Pacific Islander Alone F = Some Other Race Alone G = Two or More Races H = White Alone, Not Hispanic or Latino I = Hispanic or Latino Table Number Trivia For select tables • Final alphabetic suffix "PR" = a table used for Puerto Rico geographies only • For some geography-based subjects, the wording of the Puerto Rico Community Survey questionnaire differs slightly but significantly from the ACS questionnaire.