Adding a Mode as a Final Step in the Follow-up of a Panel Survey Seminar on New Frontiers for Statistical Data Collection Geneva, Switzerland, 31 October.

Adding a Mode as a Final Step in the Follow-up of a

Panel Survey

Seminar on New Frontiers for Statistical Data Collection Geneva, Switzerland, 31 October – 2 November 2012

Birgitta Göransson, Statistics SwedenAnton Johansson, Statistics Sweden

About the Survey• Level of Living Survey (LNU)• The survey is commisioned by Stockholm

University since 1968• Panel survey • 10 years between each wave

• Statistics Sweden has been involved in data collection in all waves

• The most recent wave of the survey began in 2010 and was finished in 2011

About the Interview• Face-to-face • Telephone as an option• Interview length approximately 60 minutes• Household survey • Data collection from several members of the

respondent´s household• Main respondent• Husband/wife/partner• Children between 10-18 years old

• Here we focus on the interview with the main respondent

Population and Sample

• Target population: Residents in Sweden, 18-74 years old

• Sample size: 8889 persons

SampleNumber of

elements

Panel 5 555

New Sample Units: interviewed as a child in LNU 2000 1 485

New Sample Units: Young adults (18-28 years) 1 381

New Sample Units: Immigrants 468

Total 8889

Response Rates in LNU 1968-2010

90 % 85% 82% 79% 76%

61%

About the Data Collection• Response rate 2010 was very low compared to the

response rate in year 2000

• But still, to reach this response rate (61 %) we had put much effort in the data collection:• For example:• Follow-ups of refusals

(in some cases more than two times)• Using experienced interviewers in the follow-up of

refusals• Tracing phone numbers and new adresses of

noncontacts

• Could anything more be done to increase the response rate?

Was it possible to do a third follow-up?• Respondent burden – ”I’ve told you NO two times

before – why can´t you understand what I’m saying?”

• Interviewer burden – could we really assign this task to the interviewers?

• We thought that the only option at this stage was to offer the nonrespondents a very short interview

• Therefore, the interview was shortened by Stockholm University to approximately 5 minutes

The Additional Mode• A short interview would most certainly have a

positive effect on the response rate

• But, couldn´t we also give the respondents the possibility to answer the questions in an optional mode?

• We wanted to try this approach.

• Therefore, in the final follow-up, we added a paper questionnaire with the same questions as in the short telephone interview

Refusals and Noncontacts

n %

Respondents 5334 61,3

- Face-to-face 3953 45,4

- Telephone 1381 15,9

Nonrespondents 3371 38,7

- Not able to participate 204 2,3

- Noncontacts 908 10,4

- Refusals 2259 26,0

Ineligible Units 184  

Sample 8889  Refusals were given the option to

answer the questions in a telephone interview or as a paper

questionnaire

Noncontacts were only given the option to answer the paper

questionnaire

Results of the Final Follow-up

90 %

72%

85% 82% 79% 76%

61%

Results of the Final Follow-up

• About half of the additional responses were paper questionnaires

• Response rate before follow-up 61,3 %

• Short telephone interviews (earlier refusals) 5,2 % • Paper questionnaires (earlier refusals) 4,0 %• Paper questionnaires (earlier noncontacts) 1,3 %

• Response rate after follow-up 71,9 %

Further work is needed

• The main concern with the follow-up was to increase the response rate

• No evaluation of mode effects with a paper questionnaire

• No evaluation if bias changed in the final follow-up

• It would be interesting to study this further!

Response Rate in the Different Sample Groups (n=8889)

  Before (%) After (%) Difference

Panel 63,1 73,9 10,8

New sample units: interviewed as a child in LNU2000

64,2 73,9 9,7

New sample units: Young Adults

55,2 67,0 11,8

New sample units: Immigrants

47,0 56,2 9,2

Total 61,3 71,9 10,6

Response Rate (by Gender)Response

rate before follow-up

Additional responses in the follow-up

Additional Responses (by Mode and Gender)

Response Rate (by Age)

Additional Responses (by Mode and Age)

Response Rate (by Education)

Additional Responses (by Mode and Education)

THANK YOU FOR YOUR ATTENTION!