MCRDC

1
MCRDC

• Michigan Census Research Data Center
• The MCRDC is a joint project of the U.S. Bureau
  of the Census and the University of Michigan to
  enable qualified researchers with approved
  projects to access confidential, unpublished
  Census Bureau data

MCRDC on the web http//www.isr.umich.edu/src/mcr
dc/
2
Agenda

• What is an RDC?
• What data are available in the RDC?
• What is the process for getting access to RDC
  data?

3
Purpose of Census Research Data Centers

• Access to data
• Secure facility
• Presence of Census Bureau employee
• Benefits to Census Bureau
• Necessary for access to Title 13 and Title 26
  data
• Not required for NCHS, AHRQ data if not linked to
  Title 13 data

4
RDCs around the country

• Ann Arbor, Michigan
• Boston, Mass.
• UCLA and Berkeley
• Baruch (NYC) and Ithaca (Cornell)
• Research Triangle, NC (Duke)
• Chicago, Illinois
• Washington, DC
• Center for Economic Studies, U.S. Census

5
Why use MCRDC data

• Not available elsewhere
• Establishment level business data
• Linked household-firm (LEHD) data
• More detail than anywhere else
• Detailed geo-spatial variables
• Virtually no top or bottom coding
• Possible to link to other non-Census data

6
Disclosure Issues

• All output goes through disclosure review process
• Disclosed output must be at state level
• Exceptions must be approved by Census review board

7
Data available in Census Research Data Centers
8
Decennial Census 1970, 1980, 1990 2000

• Comparison to PUMS
• Lowest level of geography available in the PUMS
  is an area that is roughly equivalent to a county
  but contains 100,000 people (PUMA)
• RDC version includes more detailed geographic
  information
• current residence block level
• place of work census defined place
• prior place of residence place
• This allows for aggregation to nonstandard units

9
Decennial Census

• Comparison to PUMS
• Larger sample size
• 100 of short form respondents
• One in six answered long form
• PUMS has 5 of population
• Improves analysis of small populations/sample
  sizes
• Less top-coding
• Continuous variables, such as income, are
  top-coded at a higher level
• More detailed codes (race, education, multi-race,
  e.g. type of native American)

10
American Community Survey

• All surveys with all information collected on
  survey
• Household or person-level data
• Detailed geography (census block)
• No top or bottom coding
• 1996 through 2005 currently available
• Can be linked to other data sources, where
  feasible and permissible

11
Longitudinal Business Database

• Longitudinally linked business Censuses
• All non-farm establishments with paid employees
  in (almost) all industries
• 24 million unique establishments
• Excludes airlines, agriculture, RR
• Every five years from 1977-2002
• Manufacturing Census available from 1967-2005
• Annual Survey of Manufactures includes all large
  firms

12
Longitudinal Business Database

• LBD includes
• Payroll
• Employment
• Ownership
• Detailed geographic information
• Industry at 6-digit NAICS (more detail in some
  cases)
• Other variables available (e.g. sales) but
  coverage varies across sectors

13
LBD draws on economic censuses

• Census of Manufactures
• Census of Services
• Census of Mining
• Census of Retail Trade
• Census of Wholesale Trade
• Census of Transportation, Communications and
  Utilities
• All of these Censuses are available in full, and
  can be linked over time using the LBD

14
Employer-Employee Linked Datasets

• LEHD Longitudinal Employer Household
  Dynamics
• 4 million persons linked to 1 million
  establishments
• Quarterly data on employment and wages from state
  unemployment insurance agencies
• Contains basic demographic data for all employees
• Establishments linked to the LBD
• About ½ the states are currently participating

15
Confidential Versions of Your Favorite Public Use
Datasets

• Survey of Income and Program Participation (SIPP)
  
• National Longitudinal Survey
• Current Population Survey (March)
• American Housing Survey
• National Longitudinal Mortality Survey
• Detailed geographical information
• Ability to link to other individual-level data
• Little or no top and bottom coding
• Raw data without imputations

16
Other Firm-level Datasets

• Survey of Manufacturing Technology
• Quarterly Financial Report
• US mining, manufacturing and transportation
  businesses
• Survey of Plant Capacity Utilization
• Capital Expenditure Survey
• Compustat-LBD Bridge
• National Employer Survey
• Survey of Pollution Abatement Costs and
  Expenditures
• Manufacturing Energy Consumption Survey

17
National Center for Health Statistics

• We are now hosting research using confidential
  NCHS and AHRQ data in the MCRDC
• Rules for access and disclosure the same as those
  in their enclaves
• http//www.cdc.gov/nchs/rd/rdc.htm
• http//www.meps.ahrq.gov
• No requirement to demonstrate Census benefit

18
National Center for Health Statistics

• http//www.cdc.gov/nchs/rd/rdc.htm
• National Health Interview Survey
• 1987-2004
• Some earlier years may be available
• Cross Sectional
• 43,000 Households/106,000 Persons per Year

19
National Center for Health Statistics

• http//www.cdc.gov/nchs/rd/rdc.htm
• National Survey of Family Growth
• 1995 Five Year Retrospective
• 10,847 Female Respondents
• Detailed Information on Pregnancy and
  Childbearing

20
National Center for Health Statistics

• http//www.cdc.gov/nchs/rd/rdc.htm
• Third National Health and Nutrition
  Examination Survey
• 1988-1994
• 20,000 Households/40,600 Persons
• Over Samples Young Children and Older Population

21
National Center for Health Statistics

• http//www.cdc.gov/nchs/rd/rdc.htm
• National Health Care Surveys
• National Immunization Survey
• Longitudinal Studies of Aging

22
Agency for Healthcare Research and Quality

• http//www.cdc.gov/nchs/rd/rdc.htm
• Medical Expenditure Panel Survey
• Household Component-Insurance Component linked
  file (19961999, 2001)
• Nursing Home Component (1996)
• Medical Provider Component (except directly
  identifiable data)
• Two-Year, Two-Panel Files
• Area Resource File (county-level data that can be
  linked to MEPS-HC)

23
Proposal Process

• Preliminary proposal
• Create account at Census
• http//www.ces.census.gov/
• On line submission
• Briefly describe topic and data requested
• Full proposal to MCRDC and Census (15 pgs)
• Must work with MCRDC in preparing proposal
• Must include benefits to Census as predominant
  purpose (unless asking for NCHS or AHRQ data)
• Reviewed by Census Bureaus researchers and
  external researchers

24
Virtual RDC

• Provides replica computing environment to that in
  Census RDCs
• Zero Observation data sets available
• Available variable
• Data set structure
• Ability to write code outside RDC and while
  awaiting approval of proposal
• Data sets available include LEHD, LBD, AHS
• Available at http//rdc.ciser.cornell.edu/news/

25
Census Benefits

• The predominant purpose of every project using
  Title 13 data must be to provide benefits to the
  US Census Bureau
• Proposals include an explicit statement of the
  benefits to be derived from the project
• Projects provide the Census Bureau with a
  benefits report at the completion of the
  project
• We will help you to figure out how your project
  benefits Census!

26
RDC Administrator

• Assistance in Proposal Writing
• Assistance in understanding data resources and
  what is available within a certain dataset
  (survey instruments)
• Contacts at Center for Economic Studies and
  Census program areas to help answer questions
• Lab Assistance and Security at RDC

27
Funding

• The MCRDC is supported by
• National Science Foundation
• Institute for Social Research, including the
  Survey Research Center, the Population Studies
  Center, and the Inter-University Consortium for
  Political and Social Research
• Non-UofM researchers pay 15,000 per year for
  access to Census RDCs
• Support from the Provosts office gives access at
  no charge to graduate students and faculty at the
  University of Michigan

28
Contact Information

• RDC web site http//www.isr.umich.edu/src/mcrdc/
  
• email mcrdc_at_umich.edu
• RDC phone [phone]
• RDC administrator Clint Carter
• RDC executive director Margaret Levenstein

29
Minority Women Business Owners

• SMOBE/SWOBE
• Begun in the 1960s
• Supplemented the Economic Censuses with a little
  bit of demographic information
• Characteristics of Business Owners
• 1982, 1987, 1992, 1994
• 100,000 owners surveyed
• Over-sample of women minorities
• Details on business operations and outcomes
• 1997 SMOBE/SWOBE type info only
• 2002 Survey of Business Owners
• More detail on business operations and outcomes

30
National Crime Victimization Survey

• Measures frequency, characteristics and
  consequences of criminal victimization
• Data are obtained from a national sample of
  roughly 50,000 households comprising nearly
  100,000 persons
• Differences between public use and non public
  use income not top-coded and geographic level
  more detailed
• Geographic level is Census Tract, which is a
  statistical division of a county
• 25 of sample is missing all geographic
  identifiers

31
Issues in linking data

• Is the merger feasible?
• Identifiers, statistical matching
• Business Register contains name, street, firm
  identifiers so researcher can perform matching
• Matches to individuals or households in the
  Decennial Census are performed inside Census
• Matches to geographic area characteristics can be
  done by researcher in RDC
• Is the merger permissible?
• User supplied data becomes Census Bureau data
  once co-mingled
• Must have permission from data provider