170 likes | 400 Views
GEM 2009 Adult Population Survey (APS) Review Yana Litovsky Jeff Seaman Universidad del Desarrollo, Chile January 2010. Most Important. aps@gemconsortium.org. GEM 2009 Overview. 55 Countries 183,000 + individual interviews. Multiple languages.
E N D
GEM 2009 Adult Population Survey (APS) Review Yana Litovsky Jeff Seaman Universidad del Desarrollo, Chile January 2010
Most Important aps@gemconsortium.org
GEM 2009 Overview • 55 Countries • 183,000 + individual interviews. • Multiple languages. • Each country responds to the RFP, selects their own vendor, manages their own process. • Common set of procedures and objectives. • Data harmonized to create consistent coding. • Multiple quality checks run on resulting data files.
Data Processing 2009 • RFPs submitted, checked and approved. • If a revision was required, teams resubmitted. • Only 4 countries did not require any revisions. • Typical issues: 1) insufficient number of call-backs, 2) insufficient sample size, 3) no accounting for double counting fixed line and mobile phone respondents. • Teams asked to explain how they would fix problems from previous years: 1) no random selection, 2) over-representation of females • Teams asked to fill in missing information
Data Processing 2009 • New/returning teams submitted pilot studies. • Data quality checked before being processed. • Check for missing data, incorrect skip logic, unexpected values, improper use of SPSS template and coding. • Open Ended responses coded; Education variable recoded into harmonized GEM variables. • Data processed and GEM variables/indices calculated (like TEA).
Data Processing 2009: Positives • Teams communicated well with the coordination team • fast response to emails, willingness to ask questions and follow advice, providing valuable feedback on the APS process and our work • Most teams submitted data in the proper SPSS template using appropriate variables names and codes. • Almost every new (and some returning) country submitted a pilot sample. This is also a strong suggestion for ‘veteran’ teams that change vendor.
Deadlines & Data Submission • Too many countries were late (29 countries submitted the APS over 10 days late, 6 as late as September). This produces much more work. • Over 10 teams did not submit a completed Fieldwork Report. • If this fieldwork report data is available, these need to submit it as soon as possible. • Some teams did not submit the RFP (Methodology overview) in the requested Excel template.
Data Formats • Several countries used incorrect value codes for the OCCU demographic variable. • Most countries had errors in the SPSS template. • A number of teams failed to translate certain parts into English • Some countries did not create well-distributed income ranges. • Some teams created EDUC which didn’t correspond easily to the UN or GEM education variable. • Some Open Ended responses were ambiguous, lacked detail and used local terminology.
Data Quality • All submitted datasets were inspected to make sure that the results made sense based on previous years or general expectations. • Some odd frequencies were found due to miscoded variables. • One country had to entirely resample due to oddly high entrepreneurship rates (TEA). Coordination Team worked with National Team to hypothesize source of the error and decide on a resampling strategy. Resample was successful.
Data Quality Tests • Sample Size – Does the sample meet minimum GEM requirements? • Missing or refused for key questions – What percentage of respondents did not answer critical questions? • Random assignments – Was the process of randomly assigning respondents to sets of questions done correctly? • Incomplete Interviews – What percentage of all interviews started were not completed? • Refused Interviews – What percentage refused to respond to the survey? • Gender Ratio – Does the Female/Male ratio match that of the overall population?
Gender Ratio • How well does the resulting sample match the gender distribution of the country for the age range being surveyed (18-64 or 18-99)?
Random Assignment • Are the random assignments for questions (1g to 1j) and (1k to 1n) done correctly – or is the age and gender distribution significantly different between the two groups?
Special Topic (Social Entrepreneurship) • A few teams expressed concern over the length and size of the Special Section this year. • Special Topic was long for very few respondents since most questions did not apply to most people. • One country was excused from this section due to restrictions imposed by sponsor.
Improvements made during 2009 APS cycle • GEM weights were calculated using regional population distribution for 23 countries. • APS processing syntaxes were provided with APS results to create greater transparency about our methods and allow teams to check/process their own data.
2009 Conclusions • The process from the RFP submission to the posting of the final processed datasets was smooth and communicative. • Most problems with the submitted data were small and did not affect data quality. • Still, more attention needs to be paid to details involved in data submission • All data needs to be submitted (Fieldwork Reports) and submitted on deadline.
Suggestions for Improvement • More teams to send us interim data (100 or so respondents) so that problems with data collection or entry can be identified early. • Any deviation from SPSS template and demographic variables (education) should be sent to data manager before data collection begins.
Remember aps@gemconsortium.org