The Census Bureau has posted two errata notices on AGE in ACS PUMS (# 50 (Dec 2009)  and # 47 (April 2009)).  
http://www.census.gov/acs/www/UseData/Errata.htm
 
Another paper examined PUMA level age/sex, race and Hispanic origin was presented at 2009 JSM.
"Evaluating the Age Dimension at the PUMA level in the 3-year Estimates from the ACS"
Joseph J. Salvo, Arun Peter Lobo, Adam L Willett, and Joel A. Alvarez
New York City Department of City Planning
 
http://www.amstat.org/meetings/JSM/2009/onlineprogram/index.cfm?fuseaction=abstract_details&abstractid=303441 
 
They found (in comparisons based on independent estimates using a cohort-component model (fertility, mortality, 1990-2000 migration rates)
skewed sex ratios for the age group 0-4, and overestimation of the total population of population over 65. 
"As a reults, the ACS estimates (for age 65 and over) are significantly higher than DCP projections- by 15 percent for males and 12 percent for females."
 
The last paragraph in the paper reads:
 
"Producing sub-county estimates by age/sex, race and Hispanic origin would be a major undertaking, especially in light of the recent past, when county estimates were sometimes questionable.  But this formidable challenge needs to be met if the ACS is to be a true replacement for the decennial census long form."
 
Elaine

From: ctpp-news-bounces@chrispy.net [mailto:ctpp-news-bounces@chrispy.net] On Behalf Of Krishnan Viswanathan
Sent: Wednesday, February 03, 2010 5:42 AM
To: ctpp-news@chrispy.net
Subject: [CTPP] Inaccurate age and sex data in the Census PUMS files:Evidence and Implications

This paper might be of interest to people on this list. The paper is here: http://bpp.wharton.upenn.edu/betseys/papers/Inaccurate%20Age%20and%20Sex%20Data%20in%20Census%20PUMS%20Files.pdf

From the abstract:

"We discover and document errors in public use microdata samples ("PUMS files") of the 2000 Census, the 2003-2006 American Community Survey, and the 2004-2009 Current Population Survey. For women and men ages 65 and older, age- and sex-specific population estimates generated from the PUMS files differ by as much as 15% from counts in published data tables. Moreover, an analysis of labor force participation and marriage rates suggests the PUMS samples are not representative of the population at individual ages for those ages 65 and over. PUMS files substantially underestimate labor force participation of those near retirement ages and overestimate labor force participation rates of those at older ages. These problems were an unintentional by-product of the misapplication of a newer generation of disclosure avoidance procedures carried out on the data. The resulting errors in the public use data could significantly impact studies of people ages 65 and older, particularly analyses of variables that are expected to change by age."



--
Krishnan Viswanathan
1101 High Meadow Dr
Tallahassee FL 32311