joanne pascale and nancy bates di seminar, university of essex september 17, 2004

21
Dependent Interviewing in the U.S. Census Bureau’s Survey of Income and Program Participation (SIPP) Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

Upload: jewell

Post on 14-Jan-2016

25 views

Category:

Documents


1 download

DESCRIPTION

Dependent Interviewing in the U.S. Census Bureau’s Survey of Income and Program Participation (SIPP). Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004. SIPP Overview. Multi-panel survey begun in 1983 4 month reference period; 9-12 waves per panel - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

Dependent Interviewing in the

U.S. Census Bureau’s Survey of Income and Program

Participation (SIPP)

Joanne Pascale and Nancy BatesDI Seminar, University of Essex

September 17, 2004

Page 2: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

2

SIPP Overview

• Multi-panel survey begun in 1983

• 4 month reference period; 9-12 waves per panel

• Measures dynamics of economic well-being among U.S. households

• Core topics each wave: demographics, work, government assistance, assets, health insurance

• Topical modules vary by wave

• Complex CAPI instrument

• 2004 Panel Wave 1: Sample size: 62,000; Response rate: 85.1%

Page 3: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

3

SIPP Methods Panel• 4-year comprehensive research vehicle

separate from but parallel to production SIPP survey

• General goal: reduce burden, improve response

• Dependent interviewing objectives:– Improve interview process, flow– Reduce attrition– Reduce item nonresponse– Reduce “seam” bias

Page 4: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

4

SIPP interview months and their associated reference period months (example):

OCT NOV DEC JAN FEB MAR APR MAY JUN JUL AUG

M1 M2 M3 M4 M5 Wave 1

M1 M2 M3 M4 M5 Wave 2

M1 M2 [etc.

]

Page 5: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

5

Methods Panel Design

• Expert review, analysis of existing SIPP data

• Literature review

• Qualitative research (cognitive testing, debriefing)

• 3 split-ballot field tests over 3 years

• 1st test = Wave 1; 2nd, 3rd tests = Waves 1 and 2

• 2 treatments: control and test

• Interviewers administered both treatments

• Interviewer surveys, assessments

• ~1,000 households per treatment

• Response rate = ~86%

Page 6: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

6

Dependent Interviewing Research

• Paucity of evidence in literature • Research steps:

– Commissioned formal literature review (Mathiowetz & McGonagle, 2000)

– Conducted qualitative research (Pascale & Mayer, 2003)

– Developed dependent interview methods for MPSIPP 2001 and 2002 quantitative field tests

Page 7: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

7

Types of Income Data in SIPP

• Sources of income– Jobs & businesses

– Unearned income, e.g.:• Government assistance (e.g.: food stamps, welfare)

• Government entitlements (e.g.: social security)

• Other (e.g.: child support, alimony)

– Ownership of income-producing assets, e.g.:• Stocks & bonds

• Savings accounts

• Interest-earning checking accounts

• Amount of income received from each source

Page 8: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

8

Dependent Interviewing (DI) Techniques

• Proactive: Remind and ask:‘Last time you said you received X. Do you still receive X?”

• Reactive: Ask ‘from scratch,’ offer data from prior wave if nonresponse, or to detect/resolve discrepancies

Page 9: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

9

Control DI Techniques

Remind & Re-ask W1 Q:• If reported X in W1:

‘Last time I recorded that you received/owned X. Did you receive/own X at any time between [Month 1] and today?’

• If not reported in W1: simply re-ask W1 Q

Used for unearned income and asset sources

Page 10: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

10

Control DI Techniques

Burden reduction

If R reported a job/biz in W1 ask if anything about R’s duties has changed.

If no, skip detailed questions about R’s duties and characteristics of the job/biz.

Page 11: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

11

Control DI Techniques

None• Unearned income amounts (except 2 items)• Asset amounts• Job earnings• School enrollment• Health insurance

Page 12: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

12

Control vs. Test

• DI used only for sources in control; used for both sources and amounts in test

• For sources– Control uses DI if ownership/receipt was reported

at any time during Wave 1– Test only uses DI if ownership/receipt was

reported in Month4/5 of Wave 1

• Month 5: Control does not collect ‘Month 5’ data in Wave 1; Test does

Page 13: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

13

Test DI Techniques

Remind & ask ‘still?’:

‘Last time I recorded you received X. Do you still receive X?’

Used for asset sources, health insurance

Page 14: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

14

Test DI Techniques

If source reported in M4 but not M5 of W1:Remind & ask ‘continue?’:‘Last time I recorded you received X. Did you

continue to receive X after M1 1st [of W2]?’– If yes, ask which months received– If no, probe to verify receipt really ended in M4 of

W1:• “I’m sorry to bother you about this, but it’s important for

me to get the dates just right. Are you sure X didn’t continue into [M1 of W2]?”

Page 15: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

15

Test DI Techniques

If source reported in M5:Remind & ask ‘correct?’:‘Last time I recorded you received X. Is that

correct?’ – if yes: ‘What months received in W2?’– if no: ‘Did you receive X since [M2 of W2] 1st?’

• If yes: ‘What months received in W2?’• If no: Probe to verify receipt really ended in M4 of W1

Page 16: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

16

Test DI Techniques

Ask W1 Q ‘from scratch’

If D/R: • Bring back amount provided in W1 and ask

“still sound about right?”

Used for unearned income, asset, job earnings amounts

Page 17: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

17

Test DI Techniques

Ask W1 Q ‘from scratch’

If ‘Whatever I said last time’:• Bring back amount provided in W1 and ask

‘still sound about right?’

Used for unearned income, asset, job earnings amounts

Page 18: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

18

Test DI Techniques

Corrective (mainly to catch keying errors)

• If reported in W1 re-ask W1 Q ‘from scratch’: – If new amount varies by +/-10% from W1

(unearned income amounts) or– If new amount varies by 10 fold from W1 (assets

amounts) ==> probe with interviewer note to confirm change

Page 19: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

19

Summary of DI Techniques

Topic CONTROL TEST

Jobs Burden reduction Burden reduction

Unearned income sources

Remind, re-ask W1 Depends on W1:

Remind, continue

Remind, confirm

Asset sources

Remind, re-ask W1 Remind, still

Page 20: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

20

Summary of DI Techniques

Topic CONTROL TEST

Job earnings None Nonresponse followup

Corrective

Unearned income amounts

None* Nonresponse followup

Corrective

Asset amounts None Nonresponse followup

Corrective

Health None Remind, still

Page 21: Joanne Pascale and Nancy Bates DI Seminar, University of Essex September 17, 2004

21

Next Steps

• Examine 2004 production data and compare to test results

• Assess feasibility of dependent interviewing under new windows system (vs. current DOS)