September 16th

American Community Survey (ACS)

1. American Community Survey
a. http://www.census.gov/acs/www/
b. Survey, not a census
c. ACS is potentially valuable, but in practice has much to work on
d. Results totally blurred because it has so many reference periods
e. Monthly survey
i. 2 households per 1000 per month
ii. 24 households per 1000 per year
iii. 240 households per 1000 per decade (24%)
1. Compared to 17% long form per decade
f. In principal ACS collects more data than the census, but the nonresponse rate is a lot higher for the ACS than the census
g. ACS actually has a plan to call if you don’t fill the survey out
h. If you don’t answer the call, there’s a 1 out of 3 chance they will come and visit you

2. How often the ACS collects and publishes data
a. If the population in an area is greater than 65,000, they collect data every month but publish findings once a year
i. ACS publishes this as 1 data set for a 1 year period, as if it’s a long form
ii. Average characteristics of a population over a year
iii. This poses problems because you’re taking the average now, which could skew some results. Example; locations with vacation homes where people only live in the summer.
b. If the population in an area is between 20,000-64,999 they collect data every month but wait 36 months before they publish it
i. 2006-2008 would be a data set; then 2007-2009 would be a data set; and 2008-2010 would be a data set
ii. data published as if it was a single snapshot
iii. Becomes a problem if you want to know how many kindergarteners will be attending school next year, but since findings only tell you the average from the last 3 years you don’t know how many teachers to hire
c. If the population is less than 20,000 only publish data every 60 months (including tracts, block groups, etc in bigger places)
i. Will publish data from 2006-2010; then 2007-2011; and so on 2008-2012

3. Some differences between the ACS and Census
a. The definition for residence is different in the ACS, meaning you can’t compare census resident population with ACS
b. ACS wants you to include anyone living with you two months out of the year
c. ACS has one year ago migration question, as opposed to a 5 year ago migration question in the census
d. Publish margins of errors in ACS, which census doesn’t do

Opened “ACS Questionaire” in Library

Opened “Putnam County PDF” in Course Library

1. Putnam County PDF
a. By looking at the data, we can tell the information doesn’t really tell us anything. Symmetric concept intervals are used even though the number is a negative, which can’t be less than zero since there can’t be a negative number of 20 year old females. This wouldn’t even pass a 'laugh test'. Information could potentially be useful if you could figure out a clever way to utilize it

Opened “Putnam ACS Hispanic” in Course library

1. Putnam ACS Hispanic
a. Are not able to view the noncollapsed version
b. Look at collapsed version
c. Again, shows us that data is useless. For example lists Hispanic kindergarteners: 229 = + or – 201

Can also view ACS Survey on the American Fact Finder Website

Open “ACS Summary Project” in Course library

1. ACS Summary Project
a. Basically we use ACS to estimate what people are like. For example; what percent have long commutes to work, what percent are poor, what percent are receiving government assistance
b. Do not use ACS for numbers, which isn’t logical
c. Does not tell you what a period estimate implies
d. Once a year ACS is actually useful, but you have to be skeptical about it

Public Use Microdata Sample
o What you use when they don’t have the detailed table you want
o Sample of the actual forms, which you can tabulate anyway you want
o There have been some big problems with the weighted age and sex distributions in some of the PUMS samples. ...Read about them here.
o MICROdata means individual sample

1. Go to http://ipums.org
a. Has USA and international data, USA data is not just from one time period, you can grab multiple time periods
b. Can get CPS data as well as census data
c. Can compare USA with other countries
d. Requires registration and permission, if you want international data you have to write a 250 word abstract as to why you want the info
e. Example of Data
i. Look at 2007 ACS
ii. Use variable selection, you can look at only people who have a fridge, or view people who don’t have a fridge
iii. basically you can pick choose what variables you want to look at, what cases you want to study, what years you want to look at
iv. Your requested info can be viewed in stata, spss or stats
v. Nicely organized, easy to use
vi. Includes ACS and census samples, and different subsets of long form data

page revision: 5, last edited: 18 Sep 2010 21:42