320 likes | 464 Views
Adding New Data Elements to an Ongoing Survey: Operational and Response Issues. International Conference on Establishment Surveys Montreal, 2007 Richard Rosen, U.S. Bureau of Labor Statistics rosen.richard@bls.gov
E N D
Adding New Data Elements to an Ongoing Survey: Operational and Response Issues International Conference on Establishment Surveys Montreal, 2007 Richard Rosen, U.S. Bureau of Labor Statistics rosen.richard@bls.gov Any opinions expressed in this paper are those of the authors and do not constitute policy of the Bureau of Labor Statistics.
Current Employment Statistics Survey • One of the oldest and largest US Surveys • Number of workers, payroll and hours • Monthly • Quick turnaround: • Reference period is the pay period of the 12th • 12 collection days • Data published first Friday of following month • Data closely watched by policymakers, financial markets, media, business, labor
Current Data Items • All Employees • Woman Employees • Production/Construction/Nonsupervisory • Employees • Payroll • Hours • Overtime (Manufacturing)
New Data Items • All Employee • Payroll • Hours • Overtime (Manufacturing) • Gross Monthly Earnings (GME)
Reason for New Data Items • Production workers only comprise about 80% of total, therefore the hours/earnings measures are not comprehensive • Research showed that more respondents could provide total payroll and hours • GME: broader measure of earning wanted by BEA for personal income and GDP
Data Item Issues: OLD • Production/Construction/Nonsupervisory • Definition not clearly understood • Hard to obtain from payroll records • Data item response about 40% for payroll/hours
Data Item Issues: NEW • Availability of Hours for Supervisors/managers, etc. • Multiple payrolls (i.e.. Managers paid biweekly, Nonsupervisory paid weekly) • Gross Monthly Earnings: New concept, different time period
Tasks • Forms Design • Systems Updates • Two processing/estimation systems • Six collection systems • Data transport system • Estimator for GME • Train interviewers and other staff
Planning • Began in 2003 for implementation in 2005 • Teams to work on all major aspects • Forms • Systems • Edits • Estimation
Forms • How to get additional data items on form (twice as many items) • How to handle multiple payrolls • Separation of GME from other items • Different concept/definition • Different time period (entire month) • Different “reasons for change”
Forms-Cont. • Interdisciplinary Team • Developed four different layouts • Expert review • Two rounds of cognitive field tests • Downey, Goldenberg, Rosen, Gomes, Manning, “Cognitive Testing of New Forms for the Current Employment Statistics Survey”, ASA, August 2005.
Edits: No Real History • What would relationship be between All Employee Earnings and Production Worker Earnings? • General expectation that it would be higher, but by how much • GME even more difficult • Inclusion of bonuses, stock options, severance pay could really skew data • Made some initial assumptions; revised as actual data became available
Respondents • How to notify respondents of change? • How will respondents react? • More data items • Increased burden • Will response rates suffer?
Respondent Contact • Varied by mode • Many respondents on self-reporting modes; touchtone, mail. • CATI: Interviewer began to discuss two months before change • FAX: Interviewers made special efforts to contact in advance
Respondent Contact-cont. • Special Transition Letter • Electronic: Began intensive work with firms to reprogram electronic files • Touchtone and mail: Designed “mock form” to highlight new items
Results • Overall transition has been successful • Higher data item response for All Employee Payroll/Hours • Reasonably high item response for GME But: • Overall response rates have declined
Data Item Response AE Payroll and GME reporting mostly mirrors what respondents were providing for PW Payroll by mode • TDE: Good reporters= high response • CATI/FAX: More reluctant reporter, hence need CATI= modest response • EDI: Many large firms not currently providing PW payroll; difficult to convince firms to re-program their data files; slow improvement
Looking at the new data … • Hourly Earning for All Workers averages about 15% higher than current series • Gross Monthly Earnings is somewhat volatile, shows spikes at end of each quarter, especially 4th Quarter; likely due to Bonus Payments
Conclusion • Conversion to new data elements appears to be successful. Conversion has been easier with smaller firms. • Some decline in overall response due to increased respondent burden • Trade-off in terms of improved data item response and usefulness of new data