Finding CountyBased Data from Hidden Sources

1
Finding County-Based Data from Hidden Sources

• Lisa Neidert
• Population Studies Center
• University of Michigan

2
Three Problems

• Produce county-based data from summary data
• Not all counties represented
• Produce county-based data from microdata
• County identifiers are not in microdata
• Produce county-based data from microdata
• County identifier in data
• Some county populations are too small for
  reliable data

3
American Community Survey (ACS)

• Replacement for the census long-form
  questionnaire
• 3,000,000 households a year
• County-level data every year
• Not quite

4
ACS Products Schedule
5
Distribution of US counties by size
6
Statistics based on ACS 1-year data Unit is
county
7
Statistics based on ACS 3-year data Unit is
county
8
What are PUMAs?

• Public Use Microdata areas
• Combination of population geographies that sum to
  at least 100,000 population.
• In rural areas, several counties will form a
  PUMA. In an urban area, a county will be
  subdivided into multiple PUMAs.
• PUMAs do not cross state boundaries
• Smallest geography available in the microdata.

9
Statistics based on ACS 3-year data Unit is PUMA
10
Convert PUMA-based statistics to county-based
statistics
11
PUMA-based statistic
12
Converted to county-based statistic
13
Example based on microdata

• Previous example used a table from summary data
• Distribution of the baby boom population
• Microdata allows user-generated table
• Distribution of earning equality among couples

14
Where do couples have egalitarian earnings
profiles?

• Micro-data step

15
Where do couples have egalitarian earnings
profiles?

• Micro-data step
• Produce PUMA-specific results

16
Where do couples have egalitarian earnings
profiles?

• Micro-data step
• Produce PUMA-specific results
• Convert PUMA-based results to county-based using
  cross-walk

17
What about microdata with county identifiers?

• Identifiers on Natality Detail files
• 1968-1988 all counties identified
• 1989-2005 only counties gt 100,000
• 2006 no state or county identifiers
• Distribution of births by county (1988)
• lt100 512 counties
• lt500 1,998 counties
• lt1000 2,498 counties
• Some extreme cases
• Loving county, TX 2 births
• Hinsdale county, CO 3 births
• Petroleum county, MT 3 births

18
Solution

• Cumulate small population counties by PUMA
• Calculate Fertility measures
• Total Fertility Rate
• Timing of fertility events
• Non-marital childbearing
• Use cross-walk to assign PUMA characteristic to
  counties

19
Finished Product
20
Future Directions

• Cautionary
• Pseudo-county data
• Small population-based statistics
• County population may be incorrect weight
• Web-based tool (PUMA to County)
• Input PUMA-based table
• Output County-based table
• GIS ready
• Include indicator for multi-county PUMAs