1 / 14

The Multi-Level, Multi-Source (ML-MS) Approach to Improving Survey Research

The Multi-Level, Multi-Source (ML-MS) Approach to Improving Survey Research. Tom W. Smith, NORC/University of Chicago Jibum Kim, Sungkyunkwan University. ML-MS Approach. Extract all case-level & aggregate data from sample frame Link all sample cases to auxiliary data (AD) sources

joylyn
Download Presentation

The Multi-Level, Multi-Source (ML-MS) Approach to Improving Survey Research

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. The Multi-Level, Multi-Source (ML-MS) Approach to Improving Survey Research Tom W. Smith, NORC/University of Chicago Jibum Kim, Sungkyunkwan University

  2. ML-MS Approach • Extract all case-level & aggregate data from sample frame • Link all sample cases to auxiliary data (AD) sources • Case-level • Aggregate-level • Iterative searches of AD • Paradata

  3. Sampling Frames • Available data highly variable across sampling frames and these vary across countries • Population registers; Address-based samples; Random walk • Case-level • Aggregate-levels (multi-levels)

  4. Auxiliary Data (AD) - I • Case-level links by address/other identifiers - Found/not found - If found, add additional information • Aggregate-level links - Specific address (city-style) - GPS readings - Postal code - Community and other geocodes

  5. Auxiliary Data (AD) - II • Types of AD - Household/individual-based, general databases (e.g. Accurint, Equifax, Peoplefinders) - Administrative records – single purpose databases (e.g. voting records, government benefit lists, association membership rolls)

  6. Auxiliary Data (AD) - I • Increased computerization has made more AD available and eased linkages to surveys

  7. Iterative Searches • Information gained from initial searches can facilitate additional searches • Using address-based reversed directory can find out both phone number and/or resident’s name • Name/phone number can then be used in searches of additional databases to make new links and add new information

  8. Paradata • Process • Interviewer characteristics (All) • Record of attempts (A) • Interactions with household members (A) • Keystroke (Respondents) • CARI (R) • Observational • Neighborhoods (A) • Residence, Exterior (A) • Residence, Interior (R)

  9. ML-MS Methodological Uses • Detect and adjust for non-response bias • Fill in item non-response for some variables • Validate information supplied by respondents • AD sometimes superior sources • AD other times alternative source for cross-checks • Validation interviews

  10. ML-MS Substantive Uses - I • Multi-level contextual effects • Individuals nested within households, neighborhood, communities, regions, nations • Examples of contextual, aggregate-level geographic effects from General Social Survey - racial composition of racial prejudice - crime rates on law enforcement attitudes - poverty levels and support for welfare

  11. ML-MS Substantive Uses -II • Making theory-driven, contextual, aggregate-level, geographic data a routine, automatic part of all surveys, not an special, extra (and costly) add-on

  12. Confidentially • Deductive disclosure risk increased • Need to have public file edited to insure that deductive disclosure is not possible • Use of only publically-accessible information from AD sources • Confidentiality rules vary across countries and target populations

  13. ML-MS Cross-nationally • Can be applied cross-nationally • But approach can not be standardized or rigid • Sample frames vary, available AD sources vary, and rules governing access vary • Only paradata suitable to standardization • ML-MS must adapt to those variations

  14. ML-MS Summary • Augment survey data with data from 1) sample frame, 2) AD, and 3) paradata. • Include multi-level measures from case-level and aggregate-levels (e.g. neighborhood, community) • Use for both methodological and substantive purposes • Make a routine part of doing regular surveys

More Related